@InProceedings{cnlp:aleksander94, title = "Research challenge for symbolic and neural approaches", authorkey = "AleksanderI", author = "I. Aleksander", year = "1994", number = "38", pages = "8/1--8/2", publisher = "IEE", address = "Stevenage, UK", booktitle = "IEE Colloquium on Symbolic and Neural Cognitive Engineering London, UK, Feb 14, 1994", abstract = "At general conferences papers on cognitive applications are in a small minority, and most of those are cognitive only by virtue of the fact that some world knowledge is stored as a learned function of a feed-forward neural net. This paper suggests that the future of cognitive modelling lies in a proper exploitation of neural state machines which have the power representing world knowledge and reacting appropriately to natural language. Details of this philosophy may be found in recently published literature. Here it is possible to look very briefly at some of the challenges that this work entails. 3 Refs.", keywords = "Neural networks Cognitive systems Learning systems Iconic learning", } @Article{cnlp:aleksander93, title = "Iconic neural representations and the learning of natural language", authorkey = "AleksanderI", author = "I. Aleksander", journal = "Colloquium Digest -- IEE", year = "1993", number = "92", } @InProceedings{cnlp:allen87, title = "Several Studies On Natural Language And Back-propagation.", authorkey = "AllenRB", author = "R. B. Allen", year = "1987", pages = "II/335--341", publisher = "SOS Printing", note = "Available from IEEE Service Cent (cat n 87TH0191-7), Piscataway, NJ, USA", address = "San Diego,CA", booktitle = "IEEE First International Conference on Neural Networks. San Diego, CA, June 1987", abstract = "Recent developments in neural algorithms are described that provide anapproach to natural-language processing. Two sets of brief studies show how networks may be developed for processing simple demonstratives and analogies. Two longer studies consider pronoun reference and natural-language translation. Taken together, the studies provide additional support for the applicability of these algorithms to natural language processing. 22 refs.", keywords = "Systems Science AND Cybernetics Computer Programming - Algorithms Artificial Intelligence Information Science - Language Translation and Linguistics Natural Language Processing BACK-Propagation Neural Algorithm", } @InProceedings{cnlp:allen89a, authorkey = "AllenRB RieckenME", author = "R. B. Allen and M. E. Riecken", title = "Reference in Connectionist Language Users", booktitle = "Proceedings of the International Conference on Connectionism in Perspective, Zurich, Switzerland, 11--13 Oct 1989", year = "1989", } @InProceedings{cnlp:allen89b, title = "Identifying and discriminating temporal events with connectionist language users.", authorkey = "AllenRB KaufmanSM", author = "R. B. Allen and S. M. Kaufman", year = "1989", number = "313", pages = "284--286", publisher = "IEE", ISBN = "0537-9987", address = "Stevenage, UK", booktitle = "First IEE International Conference on Artificial Neural Networks, London, UK, 16--18, Oct 1989", abstract = "The connectionist language user paradigm is applied to several studies of the perception, processing, and description of events. In one study, a network was trained to discriminate the order with which objects appeared in a microworld. In a second study, networks described sequences of events in the microworld using 'verbs.' In a third study 'plan recognition' was modeled. In the final study, networks answered questions that used verbs of possession. These results further strengthen the generality of the approach as a unified model of perception, action, and language. (author abstract) 14 Refs.", keywords = "Systems Science AND Cybernetics - Neural Nets Computer Programming Languages Automata Theory Connectionist Language Temporal Events Natural Language Sequential Network Architectures Learning Algorithms", } @Proceedings{cnlp:anon91, title = "2nd International Conference on Artificial Neural Networks.", editor = "Anon", year = "1991", number = "349", pages = "383p", publisher = "IEE", address = "Stevenage, UK", booktitle = "2nd International Conference on Artificial Neural Networks London, Bourenmouth, UK, November 1991", abstract = "This conference proceedings contains 81 papers. The main subjects are: neural networks theory, neural networks architecture, neural networks implementation, image processing applications, dynamical systems, control and robotics applications, speech and natural language processing applications, medical applications, and character recognition.", keywords = "Neural Networks Learning Systems Mathematical Techniques - Applications Statistical Methods - Applications Computer Architecture Robots, Industrial - Control Systems Eirev Supervised Learning Problem Feedforward Neural Networks Genetic Algorithms Motion Perception Statistical Pattern Recognition", } @Proceedings{cnlp:anon92a, title = "Proceedings Tenth National Conference on Artificial Intelligence", editor = "Anon", year = "1992", pages = "873p", publisher = "AAAI", address = "Menlo Park, CA", ISBN = "0-262-51063-4", booktitle = "Proceedings Tenth National Conference on Artificial Intelligence - AAAI-92, San Jose, CA, USA, July 1992", abstract = "This conference proceedings contain 132 papers. Topics covered include: Explanation and tutoring; Inductive learning, neural networks and hybrid learning, robotic learning, and learning theory; Multi-agent coordination; Natural language interpretation, and parsing; Perception; Planning; Problem solving hardness and easiness, real time problem solving, search and expert systems; Representation and reasoning of abduction and diagnosis, action and change, belief, case-based reasoning, qualitative reasoning, temporal reasoning, terminological reasoning, tractability, and qualitative model construction; Robot navigation; Scaling up; and General topics.", keywords = "Artificial intelligence Mathematical techniques Computer software Word processing Computer programming Data processing Computer applications Computation theory Algorithms Heuristic methods Computer software Learning systems Expert systems EiREV Reasoning Neural networks Constraint satisfaction problems Robot navigation Inductive learning Natural language Problem solving Representation Planning", } @Proceedings{cnlp:anon93a, title = "Speech Processing", editor = "Anon", year = "1993", volume = "2", pages = "735p", publisher = "IEEE, Piscataway, IEEE Service Center, New Jersey, USA (IEEE cat n 93CH3252-4)", ISBN = "0736-7791 0-7803-0946-4", booktitle = "1993 IEEE International Conference on Acoustics, Speech and Signal Processing Minneapolis, MN, USA Apr 27-30 1993 CC18798 IEEE Signal Processing Society", abstract = "This conference proceedings contains 189 papers from a conference on acoustics, speech and signal processing. Topics discussed include spectral quantization for speech coding, language modeling, speech recognition in noise, spoken language systems, speech coding at rates below 4 kb/s, speech synthesis, speech recognition modeling, speech recognition by neural networks, search and modeling techniques, enhancement of noisy speech, speaker and language identification, miscellaneous topics in speech coding, word spotting, speech analysis and synthesis, adaptive modeling, medium rate speech coding, and feature representation for recognition.", keywords = "Speech processing Speech coding Natural language processing systems Speech recognition Pattern recognition Speech synthesis Artificial intelligence Neural networks Signal filtering and prediction Adaptive systems Classification (of information) Random processes Algorithms Human speech processing Spectral quantization Language modeling Noisy speech enhancement Spoken language systems Speech recognition modeling Hidden Markov models Speaker recognition Word spotting", } @Proceedings{cnlp:anon94a, title = "{IEE} Colloquium on Symbolic and Neural Cognitive Engineering", editor = "Anon", organization = "IEE", year = "1994", number = "38", publisher = "IEE, Stevenage, UK", abstract = "The proceedings contains 6 papers. Some of the specific topics discussed are: planning with constraints; variable binding in a neural network using a distributed representation; and relational neurocomputing.", keywords = "Artificial intelligence Neural networks algorithms cognitive systems knowledge based systems constraint theory neurocomputing connectionist representations", } @Article{cnlp:arbitrio, title = "Towards a connectionist model of Italian morphology.", authorkey = "ArbitrioA DeloguC", author = "A. Arbitrio and C. Delogu", journal = "Neural Networks", year = "1988", volume = "1", number = "1 Suppl", pages = "287", address = "CNR, Rome, Italy", booktitle = "International Neural Network Society 1988 First Annual Meeting Boston, Massachussetts", abstract = "A lexical component for the morphological analysis of a natural language understanding system is presented. The system is implemented in Common LISP on a VAX GPX II. The lexical component is made up of two sub-systems: a segmentator for the segmentation of the input word and a lexicon containing the relations among the Italian morphemes. The lexical structure has some properties of a connectionist network formed by a set of nodes which are linked to each other and exchange activation and inhibition messages.", keywords = "Speech - Recognition Computer Programming Languages Electric Networks Summary ONLY Connectionist Model Italian Morphology Natural Language Lexical Component Segmentator", } @PhdThesis{cnlp:buchheit, authorkey = "BuchheitPJ", author = "P. J. Buchheit", title = "Infant: {A} Connectionist Like Knowledge Based and Natural Language System", school = "University of Illinois, Chicago", year = "1992", umi_no = "GAX 92-03364", } @InProceedings{cnlp:dunbar, authorkey = "DunbarG KempenM MaessenN", author = "G. Dunbar and M. Kempen and N. Maessen", title = "Semantic Interaction - {A} Connectionist Model of Lexical Combination", booktitle = "Neural Computing - Research and Applications", publisher = "IOP Publishing LTD", year = "1993", editor = "G. Orchard", pages = "95--103", address = "IOP Publishing LTD, Techno. House, Bristol BS1 6NX, UK", } @Article{cnlp:feldman, authorkey = "FeldmanJA", author = "J. A. Feldman", title = "Structured Connectionist Models and language-learning", journal = "Artificial Intelligence Review", year = "1993", volume = "7", number = "5", pages = "301--312", key_words = "Connectionist, Language Learning, Neural Network", } @Article{cnlp:harley, authorkey = "HarleyTA", author = "T. A. Harley", title = "Connectionist Approaches to Language Disorders", journal = "Aphasiology", year = "1993", volume = "7", number = "3", pages = "221--249", abstract = "This paper reviews the impact of connectionism upon our understanding of brain-damaged language performance, and attempts to explain why it is of importance for the neuropsychology of language. Connectionism is an approach to modelling cognitive processes using networks of interconnected, simple, neuron-like units. Behaviour emerges as the result of the interaction of these units. It has provided a new way of thinking about cognitive processing, emphasizing its low-level mechanisms. One supposed advantage of connectionism is its biological plausibility. It is possible to 'lesion' these systems by destroying some of the units or the connections between them. It is claimed that lesions to connectionist models of particular cognitive systems result in the appropriate acquired disorders. For example, lesioning a connectionist model of reading and word pronunciation results in surface dyslexia. Connectionist models of surface and deep dyslexia, and of word substitutions in aphasia, are described in detail. Some advantages and disadvantages of connectionism are discussed.", key_words = "Interactive Activation Model, Computational Model, Acquired Dyslexia, Letter Perception, WORD Recognition, Neural Model, Context, Paragrammatisms, Categories, Impairment", } @PhdThesis{cnlp:jain92, authorkey = "JainAN", author = "A. N. Jain", title = "Parsec: {A} Connectionist Learning Architecture for Parsing Spoken Language", school = "Carnegie Mellon University, Pittsburgh PA", year = "1992", umi_no = "GAX 92-38808", } @InCollection{cnlp:jain90, authorkey = "NJA HWA", author = "Jain A. N. and Waibel A. H.", editor = "D. S. Touretzky", title = "Incremental Parsing by Modular Recurrent Connectionist Networks", booktitle = "Advances in Neural Information Processing Systems", publisher = "Morgan Kaufman", volume = "2", pages = "364--371", year = "1990", } @Book{cnlp:kandel, editor = "A. Kandel and G. Langholz", title = "Hybrid Architectures for Intelligent Systems", publisher = "CRC Press Inc.", year = "1992", ISBN = "0-8494-4229-5", } @Article{cnlp:kwasny91a, authorkey = "KwasnySC", author = "S. C. Kwasny", title = "Connectionist Natural Language Processing", journal = "IEEE Expert-Intelligent Systems and Their Applications", year = "1991", volume = "6", number = "6", note = "Is this correct?", } @InCollection{cnlp:kwasny92a, authorkey = "KwasnySC FaisilKA", author = "S. C. Kwasny and K. A. Faisil", editor = "N. E. Sharkey", title = "A Connectionist Deterministic Parser", booktitle = "Connectionist Natural Language Processing:{ }Readings from Connection Science", publisher = "Intellect", address = "Oxford", year = "1992", } @PhdThesis{cnlp:maskara, authorkey = "MaskaraAK", author = "A. K. Maskara", title = "Recurrent Neural Networks and Grammatical Inference", school = "Polytechnic University, Brooklyn, NY", year = "1993", umi_no = "GAX 93-12245", } @Article{cnlp:ramanick, authorkey = "RamanickSG HallLO", author = "S. G. Ramanick and L. O. Hall", title = "A Hybrid Connectionist Symbolic System", journal = "Information Sciences", volume = "71", number = "3", pages = "223--268", year = "1993", } @Article{cnlp:selman, authorkey = "SelmanB", author = "B. Selman", title = "Connectionist Systems for Natural Language Understanding", journal = "Artificial Intelligence Review", year = "1989", volume = "3", number = "1", pages = "23--31", } @Article{cnlp:siegelman, authorkey = "SiegelmanHT SontagED GilesCL", author = "H. T. Siegelman and E. D. Sontag and C. L. Giles", title = "The Complexity of Language Recognition By Neural Networks", journal = "IFIP Transactions A - Computer Science and Technology", year = "1992", volume = "12", pages = "329--335", abstract = "Neural networks have been proposed as models of language acceptors. This research represents an attempt to measure the ''neural complexity'' of languages, that is, to quantify the difficulty of acceptance of a given language by a neural network. We provide a technique for estimating the number of neurons necessary to recognize a regular language using a second order neural network with various types of activation functions. We look at several different second order neural network models-with sigmoid, linear, threshold, or saturated functions-and sketch relationships between the number of neurons required in each of the models. We see that, roughly, the number in the linear activation model is an upper bound for the saturated and sigmoid models, and is a far better predictor than the minimal automaton size that had been used as an upper bound so far. Moreover, this bound is easy to compute, using techniques from the theory of rational power series in noncommuting variables.", key_words = "Formal Languages, Learning, Distributed Artificial Intelligence", } @Article{cnlp:smolensky93, authorkey = "SmolenskyP LegendreG MiyataY", author = "P. Smolensky and G. Legendre and Y. Miyata", title = "Integrating Connectionist and Symbolic Computation For The Theory Of Language", journal = "Current Science", year = "1993", volume = "64", number = "6", pages = "381--391", abstract = "In this article we present some of the fundamental principles of a research program-the Sub-Symbolic Paradigm (SSP)-based on a particular approach to unifying connectionist and symbolic computation. SSP has been developed primarily for the study of higher cognitive domains, and in this article we focus on SSP research on language and grammar. The SSP principles integrating connectionist and symbolic computation are developed by establishing mathematical relationships between two levels of description of a single computational system: at the lower level, the system is formally described in terms of highly distributed patterns of activity over connectionist units, and the dynamics of these units; at the higher level, the same system is formally described in terms of symbol structures, the constraints governing them, and the processes manipulating them. Applied to natural language, these computational principles entail that a central organizing principle of grammar is optimality: a grammar is a means of determining which of any set of structural analyses of an input is the most well-formed. Such a Harmonic Grammar consists of a set of conflicting 'soft' rules or constraints, each of which is in principle violable in the appropriate context. This constitutes a novel framework for formal grammar which emerges from the connectionist computational substrate. We describe how such soft rules allow for precise treatment of a complex set of interactions of semantic and syntactic constraints in a single language, and of universal patterns of interaction among phonological constraints.", key_words = "Simple Recurrent Networks, Distributed Representations, Neural Networks, Systems", } @InProceedings{cnlp:strain, authorkey = "StrainEP CowieRID", author = "E. P. Strain and R. I. D. Cowie", title = "Why Connectionist Networks Provide Natural Models of the Way Sentence Context Affects Identification of a Word", booktitle = "Neural Computing - Research and Applications", publisher = "IOP Publishing LTD", year = "1993", editor = "G. Orchard", pages = "85--93", address = "IOP Publishing LTD, Techno. House, Bristol BS1 6NX, UK", ISBN = "0-7503-0259-3", } @Article{cnlp:sun93a, authorkey = "SunR", author = "R. Sun", title = "Beyond Associative Memory: logic and variables in connectionist models", journal = "Information Sciences", volume = "70", number = "1--2", pages = "49--73", year = "1993", } @Article{cnlp:touretzky, authorkey = "TouretzkyDS", author = "D. S. Touretzky", title = "Special Issue on Connectionist Approaches to Language-Learning - Introduction", journal = "Machine Learning", year = "1991", volume = "7", number = "2--3", pages = "105--107", } @InProceedings{cnlp:small, authorkey = "SmallSL CottrellG ShastriL", author = "S. L. Small and G. Cottrell and L. Shastri", title = "Towards Connectionist Parsing", booktitle = "Proceedings of the National Conference on Artificial Intelligence: AAAI", year = "1982", } @InProceedings{cnlp:hinton88, authorkey = "HintonGE", author = "G. E. Hinton", title = "Representing part-whole hierarchies in connectionist networks", booktitle = "Proceedings of the 10th Annual Conference of the Cognitive Science Society, Montreal, Canada, August 1988", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", pages = "48--54", year = "1988", } @InProceedings{cnlp:pollack87a, authorkey = "PollackJB", author = "J. B. Pollack", title = "Cascaded back-propagation on dynamic connectionist networks", booktitle = "Proceedings of the 9th Annual Conference of the Cognitive Science Society", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", pages = "391--404", year = "1987", } @InCollection{cnlp:pollack89, authorkey = "PollackJB", author = "J. B. Pollack", editor = "D. S. Touretzky", title = "Implications of recursive distributed representations", booktitle = "Advances in Neural Information Processing Systems", publisher = "Morgan Kaufman", address = "San Mateo, CA", pages = "527--536", year = "1989", } @TechReport{cnlp:gasser, authorkey = "GasserME", author = "M. E. Gasser", title = "A Connectionist Model of Sequence Generation in a First and Second Language", institution = "A.I. Laboratory, Computer Science Department, University of California, Los Angeles", number = "UCLA-AI-77-13", year = "1988", } @InCollection{cnlp:sharkey, editor = "R. Trappl", authorkey = "SharkeyNE", author = "N. E. Sharkey", title = "A {PDP} System for Goal Plan Decision", booktitle = "Cybernetics and Systems", pages = "1031--1038", publisher = "Kluwer Academic", address = "Dordrecht, The Netherlands", year = "1988", } @InProceedings{cnlp:hanson, authorkey = "HansonSJ KeglJ", author = "S. J. Hanson and J. Kegl", title = "Parsnip:{A} connectionist network that learns natural language grammar from exposure to natural language sentences", booktitle = "Proceedings of the Ninth Annual Conference of the Cognitive Science Society, Seattle, WA, July 1987", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", pages = "106--119", year = "1987", } @InCollection{cnlp:villegas, authorkey = "VillegasL EbertsR", author = "L. Villegas and R. Eberts", editor = "R. Beale and J. Finlay", title = "Implementing a Neural Network for a Cognitve Text Editing Task.", booktitle = "Neural Networks and Pattern Recognition in Human-Computer Interaction", publisher = "Ellis Horwood", address = "Chichester, UK", year = "1992", ISBN = "0-13-626995-8", } @InCollection{cnlp:gorin91a, authorkey = "GorinAL LevinsonSE GertherAN GoldmanE", author = "A. L. Gorin and S. E. Levinson and A. N. Gerther and E. Goldman", title = "Adaptive Acquisition of Language", booktitle = "Neural Networks: Theory and Applications", publisher = "Academic Press Inc", year = "1991", ISBN = "0-12-467050-4", } @InProceedings{cnlp:chalmers90b, authorkey = "ChalmersDJ", author = "D. J. Chalmers", title = "Why {F}odor and {P}ylyshyn Were Wrong: The Simplest Refutation", booktitle = "Proceedings of The Twelfth Annual Conference of the Cognitive Science Society, Cambridge, MA, July 1990", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", pages = "340--347", year = "1990", } @InCollection{cnlp:chalmers92, authorkey = "ChalmersDJ", author = "D. J. Chalmers", title = "Syntactic transformations on distributed representations", booktitle = "Connectionist Natural Language Processing: Readings from Connection Science", editor = "N. E. Sharkey", publisher = "Intellect", address = "Oxford, UK", pages = "46--55", year = "1992", } @InProceedings{cnlp:niklasson, authorkey = "NiklassonL GelderT", author = "L. Niklasson and T. van Gelder", title = "Can connectionist models exhibit non-classical structure sensitivity?", booktitle = "Proceedings of the Sixteenth Annual Conference of the Cognitive Science Society, Atlanta, GA, August 1994", editor = "A. Ram", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", year = "1994", } @InCollection{cnlp:vangelder91, authorkey = "GelderT", author = "T. van Gelder", title = "What is the ``{D}'' in {PDP}? {A} survey of the concept of distribution", booktitle = "Philosophy and Connectionist Theory", editor = "W. Ramsey and S. P. Stich and D. E. Rumelhart", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", year = "1991", pages = "33--60", } @InProceedings{cnlp:hirst, authorkey = "SelmanB HirstG", author = "B. Selman and G. Hirst", title = "A rule-based connectionist parsing system", booktitle = "Proceedings of the Seventh Annual conference of the Cognitive Science Society", pages = "212--221", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", year = "1985", } @InProceedings{cnlp:jacquemin, authorkey = "JacqueminC", author = "C. Jacquemin", editor = "B. Neumann", title = "Activation Diffusion: {A} Connectionist Network for Robust Parsing", booktitle = "Proceedings of the Tenth European Conference on Artificial Intelligence(ECAI92), Vienna, Austria, August 1992", publisher = "Wiley", address = "Chichester, UK", pages = "183--187", year = "1992", } @Article{cnlp:dorffner, authorkey = "DorffnerG", author = "G. Dorffner", title = "Review on: Neural Networks for Vision Speech and Natural Language", journal = "Connection Science", volume = "5", number = "1", pages = "103--106", year = "1993", } @InProceedings{cnlp:kimura, authorkey = "KimuraL SuzuokaT AmanoS", author = "L. Kimura and T. Suzuoka and S. Amano", title = "Association Based Natural Language Processing with Neural Networks", booktitle = "30th Annual Meeting of the Association for Computational Linguistics, Newark, DE, June 1992", publisher = "Association for Computational Linguistics", pages = "224--231", year = "1992", } @Book{cnlp:Linggard, editor = "R. Linggard and D. J. Myers and C. Nightingale", title = "Neural Networks for Vision, Speech and Natural Language", publisher = "Chapman and Hall", address = "London", year = "1992", } @InCollection{cnlp:lyon92, authorkey = "LyonC FrankR", author = "C. Lyon and R. Frank", title = "Detecting Structures in Natural Language Using Neural Net and Rules", booktitle = "Artificial Neural Networks 2", editor = "I. Aleksander and J. Taylor", publisher = "North-Holland", address = "Amsterdam", pages = "731--734", year = "1992", } @InProceedings{cnlp:scholtes91e, authorkey = "ScholtesJC", author = "J. C. Scholtes", title = "Unsupervised Context Learning in Natural Language Processing", booktitle = "Proceedings of the IJCNN, Seattle, WA, July 8-12", year = "1991", volume = "1", pages = "107--112", abstract = "By generalizing over contextual information, excellent results were obtained in connectionist language processing. Normally, these contexts are added manually to the system or deducted by using a supervised learning algorithm. A recurrent self-organizing model, capable of deriving the context from scratch, is presented. It is shown that syntactic features and structures are learned in a unsupervised way from flat sentences. By generalizing over the words as well as the sentences, simple semantics can be derived. The model forms a two-layer extension of the Kohonen feature map, provided with additional recurrent fibers which are responsible for the automatic determination of word contexts, thus resulting in a unsupervised recurrent learning algorithm. After a formal description of the model, the experimental results are presented. 19 Refs.", keywords = "Learning Systems - Performance Computer Programming - Algorithms Automata Theory Natural Language Processing Learning Algorithms Semantics Syntax Kohonen Feature MAPS WORD Contexts", } @InProceedings{cnlp:scholtes91g, authorkey = "ScholtesJC", author = "J. C. Scholtes", title = "Kohonen's Self-Organizing Map Applied Towards Natural Language Processing", booktitle = "Proceedings of the CUNY 1991 Conference on Sentence Processing, Rochester, NY, May 12-14", year = "1991", pages = "10", } @Book{cnlp:hinton1990a, editor = "G. E. Hinton", title = "Connectionist Symbol Processing", publisher = "MIT Press/Elsevier", address = "Cambridge, MA", year = "1991", } @Article{cnlp:sharkey91, authorkey = "SharkeyNE", author = "N. E. Sharkey", title = "Connectionist Representation Techniques", journal = "Artificial Intelligence Review", volume = "5", number = "3", year = "1991", } @Book{cnlp:sharkey92, editor = "N. E. Sharkey", title = "Connectionist Natural Language Processing:Readings from Connection Science", publisher = "Intellect", address = "Oxford, UK", year = "1992", } @PhdThesis{cnlp:plate, authorkey = "PlateTA", author = "T. A. Plate", title = "Distributed Representations and Nested Compositional Structure", school = "University of Toronto", year = "1994", } @InProceedings{cnlp:plate91, authorkey = "PlateTA", author = "T. A. Plate", editor = "J. Mylopoulos and R. Reiter", title = "Holographic {R}educed {R}epresentations: {C}onvolution Algebra for Compositional Distributed Representations", pages = "30--35", booktitle = "Proceedings of the 12th International Joint Conference on Artificial Intelligence, Sydney, Australia, August 1991", publisher = "Morgan Kauffman", address = "San Mateo, CA", year = "1991", note = "Reprinted in Mehra P. and Wah B.W. (editors). Artificial Neural Networks: Concepts and Theory, Los Alamitos, CA, IEEE Computer Society Press 1992", } @InProceedings{cnlp:plate93a, authorkey = "PlateTA", author = "T. A. Plate", title = "Holographic Recurrent Networks", booktitle = "Advances in Neural Information Processing Systems 5: NIPS * 92, Denver, CO, November 1992", editor = "C. L. Giles and S. J. Hanson and J. D. Cowan", publisher = "Morgan Kaufmann", address = "San Mateo, CA", pages = "34--41", year = "1993", } @InProceedings{cnlp:plate94, authorkey = "PlateTA", author = "T. A. Plate", title = "Estimating structural similarity by vector dot-products of Holographic Reduced Representations", booktitle = "Advances in Neural Information Processing Systems 6 NIPS * 93, Denver, CO, November 1993", editor = "J. D. Cowan and G. Tesauro and J. Alspector", publisher = "Morgan Kaufmann", pages = "1109--1116", year = "1994", } @Article{cnlp:collier, authorkey = "CollierR", author = "R. Collier", title = "An Historical Overview of Natural Language Processing Systems that Learn", journal = "Artificial Intelligence Review", volume = "8", number = "1", year = "1994", } @TechReport{cnlp:miikkulainen93, authorkey = "MiikkulainenR", author = "R. Miikkulainen", title = "Subsymbolic Case-Role Analysis of Sentences with Embedded Clauses", institution = "University of Texas at Austin", address = "Austin, TX 78712", number = "AI 93-202", year = "1993", } @InProceedings{cnlp:nijholt, authorkey = "NijholtA", author = "A. Nijholt", editor = "R. Trappl", title = "Meta-Parsing in Neural Networks", booktitle = "Cybernetics and Systems '90", publisher = "World Scientific Publishing", address = "Singapore", pages = "969--976", year = "1990", } @InProceedings{cnlp:scholtes92, authorkey = "ScholtesJC BloembergenS", author = "J. C. Scholtes and S. Bloembergen", title = "The Design of a Neural Data-Oriented Parsing ({DOP}) System", booktitle = "IJCNN, International Joint Conference on Neural Networks", publisher = "Baltimore, IEEE", pages = "69--74", year = "1992", } @InProceedings{cnlp:lakoff, authorkey = "LakoffG", author = "G. Lakoff", title = "A Suggestion for {A} Linguistics with Connectionist Foundations", editor = "D. S. Touretzky", booktitle = "Proceedings of the 1988 Connectionist Models Summer School", publisher = "Morgan Kaufmann", address = "Los Altos,CA", pages = "301--310", year = "1988", } @InProceedings{cnlp:dorffner91, authorkey = "DorffnerG", author = "G. Dorffner", title = "'Radical' Connectionism for Natural Language Processing", booktitle = "Working Notes of the AAAI Symposium on Connectionist Natural Language Processing", year = "1991", } @Article{cnlp:jacquemin94, authorkey = "JacqueminC", author = "C. Jacquemin", title = "A Temporal Connectionist Approach to Natural Language", journal = "Sigart Bulletin", publisher = "ACM Press", volume = "5", number = "3", year = "1994", } @InProceedings{cnlp:kamimura, authorkey = "KamimuraR", author = "R. Kamimura", title = "Application of Temporal Supervised Learning Algorithm to Generation of Natural Language", booktitle = "International Joint Conference on Neural Networks, Washington DC, January 1990", publisher = "IEEE", address = "Piscataway, NJ", volume = "1", pages = "201--208", year = "1990", abstract = "An attempt is made to generate natural language by using a recurrent neural network with the temporal supervised learning algorithm (TSLA), developed by R. J. Williams and D. Zipser (1989). As TSLA uses explicit representation of consecutive events, it can deal with time-changing phenomena without increasing the number of units in the network. However, its performance has been evaluated exclusively upon the limited short sequences or sequences with explicit regularity and not for the sequences of natural language, which show complex and long-distance correlation. It was found that TSLA showed extreme instability in the learning process, and it took a long time to finish the learning. Thus, the author proposes two methods to improve the performance of TSLA. The first is the variable learning rate method, which is used to remove the instability of the learning process. The second is Minkowski-r power metrics, which is used to improve the learning time. It was found from experiments that the variable learning methods could remove the instability completely and that Minkowski-r power metrics could significantly improve the learning time. The number of hidden units, necessary for generating a long sequence of natural language, seems to reach gradually a stable point, where the number of hidden units is relatively small compared with the sequence length of natural language. This shows that TSLA can compress the information of long and complex sequences into a relatively small number of static hidden units. 9 Refs.", keywords = "Natural Language Processing Systems - Performance Neural Networks Learning Systems Learning Algorithms", } @InCollection{cnlp:sharkey90, authorkey = "SharkeyN", author = "N. Sharkey", title = "Connectionist Representation for Natural Language: Old and New", editor = "G Dorffner", booktitle = "Konnektionismus in Artificial Intelligence und Kognitionsforschung", publisher = "Springer", address = "Berlin", year = "1990", } @Book{cnlp:ward, authorkey = "WardN", author = "N. Ward", title = "A Connectionist Natural Language Generator", publisher = "Ablex", address = "Norwood, N.J.", year = "1994", note = "Revised and extended version of {\em A Flexible, Parallel Model of Natural Language Generation}, Ph.D. thesis and Technical Report UCB--CSD 91/629, Computer Science Division, University of California at Berkeley", } @Unpublished{cnlp:hammerton94, authorkey = "HammertonJA", author = "J. A. Hammerton", title = "A Hybrid Connectionist Shift Reduce Parser", note = "Note: Final Year Project Report, Departments of Artificial Intelligence and Computer Science, The University of Edinburgh", month = jun, year = "1994", } @Article{cnlp:nenov93, title = "Perceptually Grounded Language Learning: Part 1-{A} Neural Network Architecture for Robust Sequence Association", authorkey = "NenovVI DyerMG", author = "V. I. Nenov and M. G. Dyer", journal = "Connection Science", year = "1993", volume = "5", number = "2", } @Article{cnlp:nenov, title = "Perceptually Grounded Language Learning: Part 2-{DETE}: {A} Neural/Procedural Model", authorkey = "NenovVI DyerMG", author = "V. I. Nenov and M. G. Dyer", journal = "Connection Science", year = "1994", volume = "6", number = "1", ISBN = "0954-0091", } @Article{cnlp:gupta, title = "Connectionist Models and Linguistic Theory: Investigations of Stress Systems in Language", authorkey = "GuptaP TouretzkyDS", author = "P. Gupta and D. S. Touretzky", journal = "Cognitive Science", year = "1994", volume = "18", number = "1", ISBN = "0364-0213", } @Article{cnlp:sharkey94a, title = "Connectionist advances in natural language processing", authorkey = "SharkeyN", author = "N. Sharkey", journal = "Colloquium Digest -- IEE", year = "1994", number = "38", } @Article{cnlp:lyon93, title = "Using neural networks to infer grammatical structures in natural language", authorkey = "LyonC", author = "C. Lyon", journal = "Colloquium Digest -- IEE", year = "1993", number = "92", } @Article{cnlp:barnden93, title = "Connectionist meta-representation for propositional attitudes", authorkey = "BarndenJA", author = "J. A. Barnden", journal = "Journal of Experimental and Theoretical Artificial Intelligence", year = "1993", volume = "5", number = "2--3", } @Article{cnlp:surkan, title = "Coding of Natural Language Task Descriptions prior to Their Classification by Neural Networks", authorkey = "SurkanAJ EvansRM", author = "A. J. Surkan and R. M. Evans", journal = "Proceedings Of The Hawaii International Conference On System Sciences", year = "1993", volume = "26//V4", } @Article{cnlp:utsumi, title = "Natural Language Interfase using Contextual Information by Connectionist Model", authorkey = "UtsumiA HoriK OhsugaS", author = "A. Utsumi and K. Hori and S. Ohsuga", journal = "Japanese Society For Artificial Intelligence", year = "1992", volume = "7", number = "5", } @Article{cnlp:moisl, title = "Connectionist Finite State Natural Language Processing", authorkey = "MoislH", author = "H. Moisl", journal = "Connection Science", year = "1992", volume = "4", number = "2", } @InProceedings{cnlp:wu, title = "Word sense disambiguation by a higher order connectionist net based on distributed representations", authorkey = "WuX McTearM OjhaP", author = "X. Wu and M. McTear and P. Ojha", booktitle = "Proceedings of the 1993 IEEE Region 10 Conference on Computer, Communication, Control aand Power Engineering (Tencon '93). Part 2 (of 5) Beijing, China Oct 19-21 1993 CC20220 IEEE", year = "1993", pages = "893--897", ISBN = "0-7803-1233-3", address = "Univ of Ulster at Jordanstown, Ireland", abstract = "Word sense disambiguation is one of the most challenging areas in natural language processing. In this paper, we describe a higher order neural net based on a distributed representation method for word sense disambiguation. Two types of theory of neural representations, which are localist and distributed representations, are discussed and compared. The concept of microfeatures which typically belongs to the category of distributed representation is introduced and incorporated in the system. We will show that a system based on a distributed representation is potentially more effective than that based on a localist representation in solving the word sense disambiguation problem. (Author abstract) 18 Refs.", keywords = "Neural networks Word sense disambiguation Higher order connectionist net Neural representation theories _723 _921", } @Article{cnlp:chandola, title = "Ordered rules for full sentence translation: a neural network realization and a case study for {H}indi and {E}nglish", authorkey = "ChandolaA MahalanobisA", author = "A. Chandola and A. Mahalanobis", journal = "Pattern Recognition", year = "1994", volume = "27", number = "4", pages = "515--521", publisher = "Pergamon Press Inc, Tarrytown, New York, USA", ISBN = "0031-3203", address = "Univ of Arizona, Tucson, Arizona, USA", abstract = "As a general tool for pattern recognition, neural networks have made an impact in various fields which require automation and learning. Their use is demonstrated for learning ordered rules which make it possible to translate one language into another. Using pattern recognition techniques, it is shown that neural networks can implement such rules by learning to reorder the words in a sentence appropriately. First it is described how the ordered rules yield a situational translation from one language to another. Illustrative examples are given in English and Hindi, two Indo-European languages. Then the neural network realization of the translation algorithm is discussed. The work presented here may serve as a model for other languages. (author abstract) 7 Refs.", keywords = "Pattern recognition Linguistics Translation (languages) Neural networks Learning systems Algorithms Computer applications Ordered rules Full sentence translation Phrase order Situation Reordering Coded input/output Hindi", } @Article{cnlp:kita, title = "Spoken sentence recognition based on {HMM}-{LR} with hybrid language modeling", authorkey = "KitaK MorimotoT OhkuraK SagayamaS YanoY", author = "K. Kita and T. Morimoto and K. Ohkura and S. Sagayama and Y. Yano", journal = "Transactions on Information and Systems 94 VE", year = "??", number = "2", pages = "258--265", publisher = "Inst of Electronics, Inf \& Commun Engineers of Japan, Tokyo, Japan", ISBN = "0916-8532", address = "Tokushima Univ, Tokushima-shi, Japan", abstract = "This paper describes Japanese spoken sentence recognition using hybrid language modeling, which combines the advantages of both syntactic and stochastic language models. As the baseline system, we adopted the HMM-LR speech recognition system, with which we have already achieved good performance for Japanese phrase recognition tasks. Several improvements have been made to this system aimed at handling continuously spoken sentences. The first improvement is HMM training with continuous utterances as well as word utterances. In previous implementations, HMMs were trained with only word utterances. Continuous utterances are included in the HMM training data because coarticulation effects are much stronger in continuous utterances. The second improvement is the development of a sentential grammar for Japanese. The sentential grammar was created by combining inter- and intra-phrase CFG grammars, which were developed separately. The third improvement is the incorporation of stochastic linguistic knowledge, which includes stochastic CFG and a bigram model of production rules. The system was evaluated using continuously spoken sentences from a conference registration task that included approximately 750 words. We attained a sentence accuracy of 83.9% in the speaker-dependent condition. (author abstract) 23 Refs.", keywords = "Speech recognition Natural language processing systems Computer simulation Computational linguistics Neural networks Context free grammars Acoustics Speech analysis Hybrid language modeling HMM-LR speech recognition Stochastic linguistic knowledge LR sparsing Hidden Markov model Syntactic language model Acoustic information Stochastic language model", } @InProceedings{cnlp:law, title = "Neural network-assisted Japanese-English machine translation system", authorkey = "LawT ItohH SekiH", author = "T. Law and H. Itoh and H. Seki", booktitle = "Proceedings of 1993 International Joint Conference on Neural Networks. Part 3 (of 3) Nagoya, Jpn Oct 25, 1993", year = "1993", volume = "3", pages = "2905--2908", publisher = "IEEE", address = "Piscataway, NJ", ISBN = "0-7803-1421-2", abstract = "In this paper, we present a hybrid machine translation system which combines the strengths of logic programming, procedural programming, and neural networks. The system is designed for Japanese to English translation of natural language sentences found in daily newspaper weather reports. The system is trained on one full year of weather reports, or over 1000 sentences. A second full year of weather reports is used to empirically evaluate the system. (author abstract) 5 Refs.", keywords = "Computer aided language translation Linguistics Neural networks Logic programming Natural language processing systems Artificial intelligence Expert systems Systems analysis Hybrid machine translation system", } @InProceedings{cnlp:wang93, title = "Simple recurrent network for Chinese word prediction", authorkey = "WangM LiuW ZhongY", author = "M. Wang and W. Liu and Y. Zhong", year = "1993", volume = "1", pages = "263--266", publisher = "IEEE, Piscataway, New Jersey, USA (IEEE cat n 93CH3353-0)", ISBN = "0-7803-1421-2", address = "Beijing Univ, Beijing, China", booktitle = "Proceedings of 1993 International Joint Conference on Neural Networks. Part 1 (of 3) Nagoya, Jpn Oct 25-25", abstract = "This paper presents preliminary investigations concerning the use of Simple Recurrent Network (SRN) in Chinese word prediction. We explore the architecture introduced by J.L.Elman for predicting successive elements of a sequence. This model is based on a multi-layer architecture and contains special units, called context units which provide the short-term memory(STM) in the system. Based on this model, We constructed a modular SRNs to predict Chinese word at two levels. The first level network predicts the major category of the next word, then the next possible word is predicted at the second level network. Also, the specific encoding schemes was described in the paper. Experiments show that the method is promising. (author abstract) 5 Refs.", keywords = "Neural networks Word processing Computer architecture Encoding (symbols) Computer simulation Data storage equipment Natural language processing systems Simple recurrent network Chinese word prediction Short term memory Word prediction problem High level cognitive processing Time sequence", } @InProceedings{cnlp:kawahara, title = "Neural network approach to inference mechanism for logic programming language", authorkey = "KawaharaH MurakoshiH FunakuboN IshijimaS", author = "H. Kawahara and H. Murakoshi and N. Funakubo and S. Ishijima", year = "1993", volume = "1", pages = "167--170", publisher = "IEEE", address = "Piscataway, NJ", ISBN = "0-7803-1421-2", booktitle = "Proceedings of 1993 International Joint Conference on Neural Networks. Part 1 (of 3) Nagoya, Japan Oct 25--29", abstract = "We present a new inference mechanism for logic programming languages using neural networks that is flexible and suited for fine-grain parallel computing. Our approach is radically different from the conventional methods based on refutation processes. Programs written in logic programming language is transformed into a Hopfield-type neural network. And relaxation techniques are applied to this network to inference solutions. We proposed an algorithm to transform logic programs into Hopfield-type neural networks and implemented a prototype of the inference system based on this mechanism. And we tested the system with some preliminary problems. This preliminary results confirm us that our algorithm is correct. (author abstract) 4 Refs.", keywords = "Neural networks Inference engines Logic programming Parallel processing systems Algorithms Relaxation processes Optimization Computer programming languages Computational methods Logic programming language Parallel computing Refutation processes Energy function Hopfield type neural network", } @InProceedings{cnlp:tsunoda, title = "Semantic ambiguity resolution by parallel distributed associative inference and contradiction detection", authorkey = "TsunodaT TanakaH", author = "T. Tsunoda and H. Tanaka", year = "1993", volume = "1", pages = "163--166", publisher = "IEEE, Piscataway, New Jersey, USA (IEEE cat n 93CH3353-0)", ISBN = "0-7803-1421-2", address = "Univ of Tokyo, Tokyo, Japan", booktitle = "Proceedings of 1993 International Joint Conference on Neural Networks. Part 1 (of 3) Nagoya, Jpn Oct 25-29", abstract = "In this paper we propose a PDAI&CD architecture aimed at constructing natural inference systems. The kernel consists of a mutually associative neural network which processes numerical patterns and of a logical system processing symbols. The associative part calls on context-dependent free-association of concepts based on the relations of concepts acquired from a dynamically changing outer world. In the logical part of the architecture, the results obtained by the neural network are checked, and emerging contradictions create feedback to the associative network and thus find a final optimum solution. WAVE, a concrete implemented system, is introduced here, and we show its application to the problem of ambiguity resolution in natural language understanding. (author abstract) 3 Refs.", keywords = "Inference engines Neural networks Database systems Computational linguistics Data acquisition Computer architecture Codes (symbols) Parallel processing systems Natural language processing systems Context sensitive languages Semantic ambiguity resolution Natural inference systems Kernel Logical system processing Syntactics Winner associative voting engine", } @Article{cnlp:sun93b, title = "Efficient feature-based connectionist inheritance scheme", authorkey = "SunR", author = "R. Sun", journal = "IEEE Transactions on Systems, Man and Cybernetics", year = "1993", volume = "23", number = "2", pages = "512--522", ISBN = "0018-9472", address = "Univ of Alabama, Tuscaloosa, Alabama, USA", abstract = "A connectionist model that deals with the inheritance problem in an efficient and natural way is described. Based on the connectionist architecture Consyderr, the problem of property inheritance and formulate it in ways facilitating conceptual clarity and connectionist implementation is analyzed. A set of `benchmarks' is specified for ensuring the correctness of solution mechanisms; parameters of Consyderr are formally derived to satisfy these benchmark requirements. The paper also discusses how chaining of is-a links and multiple inheritance can be handled in this architecture. It is shown that Consyderr with a two-level dual (localist and distributed) representation can handle inheritance and cancellation of inheritance correctly and extremely efficiently, in constant time instead of proportional to the length of a chain in an inheritance hierarchy. It also demonstrates the utility of a meaning-oriented intentional approach (with features), for supplementing and enhancing extensional approaches. (author abstract) 31 Refs.", keywords = "Artificial intelligence Data structures Pattern recognition Algorithms Knowledge based systems Hierarchical systems Systems analysis Inference engines Fuzzy sets Mathematical models Computer architecture Efficient feature based connectionist scheme Connectionist architecture Consyderr", } @InProceedings{cnlp:kock, title = "Connectionist model description: {A} case study", authorkey = "KockG SerbedzijaNB", author = "G. Kock and N. B. Serbedzija", year = "1993", pages = "843--850", ISBN = "0165-6074", booktitle = "19th Euromicro Symposium on Microprocessing and Microprogramming, Barcelona, Spain, September 6-9, 1993", abstract = "An approach to provide high-level specification of neural networks is presented in this paper and illustrated on the well known `truck backer-upper' problem. The developed specification language MAX is based upon the modified AXON model and provides readable descriptions of common connectionist models. The descriptions contain all relevant information concerning connectionist programming but are free from implementation details. Further refinement and implementation strategy is also outlined. (author abstract) 8 Refs.", keywords = "Neural networks Computer programming Computer hardware description languages Mathematical models Control systems Computer architecture Computer simulation C (programming language) Programming theory Connectionist model description Truck backer upper problem Specification language MAX", } @InProceedings{cnlp:baldwin, title = "Connectionist vs. symbolic feature representation in evidential support logic", authorkey = "BaldwinJF CoyneMR MartinTP", author = "J. F. Baldwin and M. R. Coyne and T. P. Martin", year = "1993", pages = "827--832", publisher = "IEEE", address = "Piscataway, NJ", ISBN = "0-7803-0615-5", booktitle = "Second IEEE International Conference on Fuzzy Systems San Francisco, CA, March 28--April 1, 1993", abstract = "The use of evidential support logic in deductive reasoning is introduced. Two methods of matching features within objects are considered which are based on a symbolic and a sub-symbolic approach. The two are compared in a small example which uses four hand-written characters, training is performed over a small set of 'perfect' examples and a set of untrained test cases are considered. The conclusions suggest that although both methods have certain merits, the combination of the advantages of each are possible due to the transparency of the feature mapping method to the rest of the evidential support engine. (author abstract) 12 Refs", keywords = "Automata theory Logic design Fuzzy sets Character recognition Image processing Evidential support logic Feature representation Deductive reasoning Feature mapping", } @Article{cnlp:romaniuk, title = "{SC}-net: {A} hybrid connectionist, symbolic system", authorkey = "RomaniukSG HallLO", author = "S. G. Romaniuk and L. O. Hall", journal = "Information Sciences", year = "1993", volume = "71", number = "3", pages = "223--268", ISBN = "0020-0255", address = "Univ of South Florida, Tampa, Florida, USA", abstract = "This paper describes the SC-net system that has been developed to provide expert systems capability augmented with learning in a hybrid connectionist/symbolic approach. A distributed connectionist representation of cells connected by links is used to represent symbolic knowledge. Rules may be directly encoded in the connectionist network or learned from examples. The learning method is a form of instance-based learning in which some of the individual instances in the training set are encoded by adding structure to the network and others cause modifications to biases in the network. Both continuous and nominal attributes are directly represented in the network structure. A limited form of variables in the form of attribute value bindings on the right-hand side of rules is supported. Relational comparators in the form of cell groups are also supported. Relational comparators and attribute value structures are represented by groups of connected cells in the network. The learning algorithm is presented and methods for providing generalization in an instance-based connectionist environment are presented. Empirical results are presented, which include learning in domains (fevers and gems) that contain uncertainty and the well-known iris, and soybean data sets together with a real world domain for semiconductor wafer fault diagnosis. The generalization ability of the learned network is shown to be good in several domains including iris. The system is shown to compare favorably with a nonneural instance-based learning algorithm IBL. (author abstract) 49 Refs", keywords = "Learning systems Neural networks Expert systems Algorithms Encoding (symbols) Data structures Relational comparators Learning algorithms", } @Article{cnlp:fu, title = "Knowledge-based connectionism for revising domain theories", authorkey = "FuLM", author = "L. M. Fu", journal = "IEEE Transactions on Systems, Man and Cybernetics", year = "1993", volume = "23", number = "1", pages = "173--182", ISBN = "0018-9472", address = "Univ of Florida, Gainesville, Florida, USA", abstract = "Integration of domain theory into empirical learning is important in building a useful learning system in practical domains since the theory is not always perfect and the data is not always adequate. A novel knowledge-based connectionist model referred to as Kbcnn for machine learning is presented. In the Kbcnn learning model, useful domain attributes and concepts are first identified and linked in a way consistent with initial domain knowledge, and then the links are weighted properly so as to maintain the semantics. Hidden units and additional connections may be introduced into this initial connectionist structure as appropriate. Then, this primitive structure evolves to minimize empirical error. The Kbcnn learning model allows the theory learned or revised to be translated into the symbolic rule-based language that describes the initial theory. Thus, a domain theory can be pushed onto the network, revised empirically over time, and decoded in symbolic form. The domain of molecular genetics has been used to demonstrate the validity of the Kbcnn learning model and its superiority over related learning methods. (author abstract) 30 Refs", keywords = "Knowledge based systems Learning systems Neural networks Cognitive systems Domain theory Empirical learning Machine learning Knowledge based connectionist models Rule based systems", } @Article{cnlp:maclennan, title = "Characteristics of connectionist knowledge representation", authorkey = "MaclennanB", author = "B. Maclennan", journal = "Information Sciences", year = "1993", volume = "70", number = "1--2", pages = "119--143", ISBN = "0020-0255", address = "Univ of Tennessee, Knoxville, TE, USA", abstract = "Connectionism -- the use of neural networks for knowledge representation and inference -- has profound implications for the representation and processing of information because it provides a fundamentally new view of knowledge. However, its progress is impeded by the lack of a unifying theoretical construct corresponding to the idea of a calculus (or formal system) in traditional approaches to knowledge representation. Such a construct, called a simulacrum, is proposed here, and its basic properties are explored. We find that although exact classification is impossible, several other useful, robust kinds of classification are permitted. The representation of structured information and constituent structure are considered, and we find a basis for more flexible rule-like processing than that permitted by conventional methods. We discuss briefly logical issues such as decidability and computability and show that they require reformulation in this new context. Throughout, we discuss the implications of this new theoretical framework for artificial intelligence and cognitive science. (author abstract) 31 Refs", keywords = "Neural networks Knowledge based systems Artificial intelligence Formal logic Computability and decidability Learning systems Adaptive systems Cognitive systems Pattern recognition Combinatorial mathematics Connectionist networks Connectionist knowledge representation Flexible information processing Connectionism constructs Cognitive science", } @InProceedings{cnlp:berg, title = "Connectionist parser with recursive sentence structure and lexical disambiguation", authorkey = "BergG", author = "G. Berg", publisher = "AAAI", year = "1992", pages = "32--37", address = "Menlo Park, CA", ISBN = "0-262-51063-4", booktitle = "Proceedings of the Tenth National Conference on Artificial Intelligence - AAAI-92, San Jose, CA, July 1992", abstract = "In order to be taken seriously, connectionist natural language processing systems must be able to parse syntactically complex sentences. Current connectionist parsers either ignore structure or impose prior restrictions on the structural complexity of the sentences they can process - either number of phrases or the 'depth' of the sentence structure. Xeric networks, presented here, are distributed representation connectionist parsers which can analyze and represent syntactically varied sentences, including ones with recursive phrase structure constructs. No a priori limits are placed on the depth or length of sentences by the architecture. Xeric networks use recurrent networks to read words one at a time. RAAM-style reduced description and X-Bar grammar are used to make an economical syntactic representation scheme. This is combined with a training technique which allows Xeric to use multiple, virtual copies of its RAAM decoder network to learn to parse and represent sentence structure using gradient-descent methods. Xeric networks also perform number-person disambiguation and lexical disambiguation. Results show that the networks works train to a few percent error for sentences up a phrase-nesting depth of ten or more and that this performance generalizes well. (author abstract) 11 Refs", keywords = "Artificial intelligence Computer programming languages Computer systems programming Word processing Connectionist parser Recursive sentence structure Lexical disambiguation Xeric networks Natural language processing Phrase nesting depth RAAM decoder network.", } @InProceedings{cnlp:gorin91b, title = "On adaptive acquisition of spoken language", authorkey = "GorinAL LevinsonSE MillerLG GertnerAN", author = "A. L. Gorin and S. E. Levinson and L. G. Miller and A. N. Gertner", year = "1991", pages = "422--431", publisher = "IEEE", address = "New York, USA", ISBN = "0-7803-0118-8", booktitle = "Proceedings of the 1991 Workshop on Neural Networks for Signal Processing - NNSP-91, Princeton, NJ, September 1991", abstract = "Progress in building a device capable of acquiring the necessary linguistic skills for high-perforkance speech synthesis is reported. Some principles and mechanisms upon which such a device might be based are described. Several rudimentary experiments evaluating their utility are recounted.", keywords = "speech processing learning systems neural networks language acquisition adaptive language acquisition spoken language", } @InProceedings{cnlp:jaravine, title = "Syntactic neural network for character recognition", authorkey = "JaravineVA", author = "V. A. Jaravine", year = "1992", pages = "215--223", publisher = "Int Soc for Optical Engineering", address = "Bellingham, Washington", ISBN = "0277-786X 0-8194-0815-8", booktitle = "Machine Vision Applications in Character Recognition and Industrial Inspection, San Jose, CA, February 1992", abstract = "This article presents a synergism of syntactic 2-D parsing of images and multilayered, feed- forward network techniques. This approach makes it possible to build a written text reading system with absolute recognition rate for unambiguous text strings. The Syntactic Neural Network (SNN) is created during image parsing process by capturing the higher order statistical structure in the ensemble of input image examples. Acquired knowledge is stored in the form of hierarchical image elements dictionary and syntactic network. The number of hidden layers and neuron units is not fixed and is determined by the structural complexity of the teaching set. A proposed syntactic neuron differs from conventional numerical neuron by its symbolic input/output and usage of the dictionary for determining the output. This approach guarantees exact recognition of an image that is a combinatorial variation of the images from the training set. The system is taught to generalize and to make stochastic parsing of distorted and shifted patterns. The generalizations enables the system to perform continuous incremental optimization of its work. New image data learned by SNN doesn't interfere with previously stored knowledge, thus leading to unlimited storage capacity of the network. 8 refs.", keywords = "Character recognition Neural networks Statistical methods Digital image storage Image analysis Learning systems Syntactic neural network Feed-forward neural network Image parsing Stochastic parsing Training set Local/global dictionary Written text recognition Syntactic neuron", } @Article{cnlp:hunt, title = "Neural networks for control systems - a survey", authorkey = "HuntKJ SbarbaroD ZbikowskiR GawthropPJ", author = "K. J. Hunt and D. Sbarbaro and R. Zbikowski and P. J. Gawthrop", journal = "Automatica", year = "1992", volume = "28", number = "6", pages = "1083--1112", ISBN = "0005-1098", address = "Univ of Glasgow, Glasgow, UK", abstract = "This paper focuses on the promise of artificial neural networks in the realm of modelling, identification and control of nonlinear systems. The basic ideas and techniques of artificial neural networks are presented in language and notation familiar to control engineers. Applications of a variety of neural network architectures in control are surveyed. We explore the links between the fields of control science and neural networks in a unified presentation and identify key areas for future research. (Author abstract) 196 Refs", keywords = "Neural networks Nonlinear control systems Identification (control systems) Computer architecture Learning systems Predictive control systems Signal filtering and prediction Cellular neural networks Neuron nonlinearity Fixed point learning Trajectory learning Model reference control Adaptive linear control", } @InProceedings{cnlp:dasgupta, title = "Learning capabilities of recurrent neural networks", authorkey = "DasGuptaB", author = "B DasGupta", year = "1992", volume = "2", pages = "822--823", publisher = "IEEE", address = "Piscataway, NJ", ISBN = "0734-7502 0-7803-0494-2", booktitle = "Proceedings of the IEEE Southeastcon '92 Birmingham, AL, April 1992", abstract = "The author relates the power of recurrent neural networks to those of other conventional models of computation like Turing machines and finite automata, and proves results about their learning capabilities. Specifically, it is shown that (a) probabilistic recurrent networks and probabilistic Turing machine models are equivalent; (b) probabilistic recurrent networks with bounded error probabilities are not more powerful than deterministic finite automata; (c) deterministic recurrent networks have the capability of learning P-complete language problems; and (d) restricting the weight-threshold relationship in deterministic recurrent networks may allow the network to learn only weaker classes of languages. 4 Refs", keywords = "Neural networks Learning systems Turing machines Finite automata Probability Computer programming languages Recurrent neural networks Probabilistic neural networks Deterministic recurrent networks", } @InProceedings{cnlp:oka, title = "Hybrid cognitive model of conscious level processing and unconscious level processing", authorkey = "OkaN", author = "N. Oka", year = "1992", pages = "485--490", publisher = "IEEE, Piscataway, IEEE Service Center, New Jersey, USA (IEEE cat n 92CH3065-0)", ISBN = "0-7803-0227-3", address = "Matsushita Res Inst Tokyo, Inc, Tama-ku, Kawasaki, Japan", booktitle = "1991 IEEE International Joint Conference on Neural Networks - IJCNN '91 Singapore, Singapore", abstract = "Human intelligence has been modeled in two ways: modeling based on central symbolic processing, and modeling based on distributed subsymbolic processing. This paper points out the limitations of these two approaches, and proposes a hybrid cognitive model of central symbolic processing on the conscious level, and distributed subsymbolic processing on the unconscious level (C/U model). The advantages of the C/U model are clarified by explaining various functions realized by the model: multistage knowledge retrieval, recognition and inference with situated knowledge, inductive learning, and creative inference. Those functions are realized by utilizing the close interaction between the two levels. Finally, this paper describes an implementation method that utilizes the characteristics of a parallel logic programming language, and also describes a knowledge acquisition system necessary for building practical hybrid systems. 15 Refs", keywords = "Cognitive systems Logic programming Cognitive models Conscious level processing Unconscious level processing Distributed subsymbolic processing Knowledge acquisition", } @Article{cnlp:rocha, title = "A neural net for extracting knowledge from natural language data bases", authorkey = "RochaAF GuilhermeIR TheotoM MiyadahiraAMK KoizumiMS", author = "A. F. Rocha and I. R. Guilherme and M. Theoto and A. M. K. Miyadahira and M. S. Koizumi", journal = "IEEE Transactions on Neural Networks", year = "1992", volume = "3", number = "5", pages = "819--828", ISBN = "1045--9227", abstract = "A model of a fuzzy neuron, one which increases the computational power of the artificial neuron, turning it also into a symbolic processing device, is presented. The model proposes the synapsis to be symbolically and numerically defined, by means of the assignment of tokens to the presynaptic and postsynaptic neurons. The matching or concatenation compatibility between these tokens is used to decide about the possible connections among neurons of a given net. The strength of the compatible synapsis is made dependent on the amount of the available presynaptic and postsynaptic tokens. The symbolic and numeric processing capacity of the new fuzzy neuron is used to build a neural net (Jargon) to disclose the existing knowledge in natural language databases such as medical files, sets of interviews and reports about engineering operations. 22 Refs", keywords = "Neural networks Natural language processing systems Artificial intelligence Codes (symbols) Fuzzy sets Database systems Medical computing Knowledge extraction Natural language databases Fuzzy neurons Artificial neurons Symbolic processing", } @InProceedings{cnlp:gallant, title = "A practical approach for representing context and for performing word sense disambiguation using neural networks.", authorkey = "GallantSI", author = "S. I. Gallant", year = "1992", pages = "1007", publisher = "IEEE", address = "Piscataway, NJ", ISBN = "0-7803-0164-1", booktitle = "International Joint Conference on Neural Networks - IJCNN-91-Seattle Part 2 (of 2) Seattle, Washington 08 - 12 July 1991", abstract = "Summary form only given. The author proposes a method for representing context information so that the correct meaning for a word in a sentence can be selected. The approach is primarily based upon work by Waltz and Pollack, who emphasized neurally plausible systems. By contrast the author focuses upon computationally feasible methods applicable to full-scale natural language processing systems. There are two key elements: a collection of context vectors defined for every word used by a natural language processing system, and a context algorithm that computes a dynamic context vector at any position in a body of text. Once the dynamic context vector has been computed it is easy to choose among competing meanings for a word. This choice of definitions is essentially a neural network computation, and neural network learning algorithms should be able to improve the system's choices. Good candidates for full-scale context vector implementations are machine translation systems and Japanese word processors.", keywords = "neural networks - applications learning systems computer programming - algorithms dynamics data processing - word processing japanese word processors summary only", } @InProceedings{cnlp:iooss, title = "From lattices of phonemes to sentences: {A} recurrent neural network approach.", authorkey = "IoossC", author = "C. Iooss", year = "1992", pages = "833--838", publisher = "IEEE", address = "IEEE Service Center, Piscataway, NJ, USA.", ISBN = "0-7803-0164-1", booktitle = "International Joint Conference on Neural Networks - IJCNN-91-Seattle Part 2 (of 2) Seattle, Washington, 8--12 July 1991", abstract = "The author presents preliminary investigations concerning the use of a sequential neural network for lexical decoding in continuous speech recognition. They explore the architecture introduced by J.L. Elman (Center for Research in Language Technical Report 8801, UCSD, April 1988) for predicting successive elements of a sequence. This recurrent network admits sequential inputs. This model is based on a multilayer architecture and contains special units, called context units, sensitive to the recent activation history of the network. It is suggested that this model be used for lexical decoding in continuous speech recognition. For that purpose, an extension of Elman's model is presented in order to treat erroneous sequential inputs and in order to label patterns. It is suggested that the context units be updated considering their previous values and not only the values of the hidden units. Moreover, output units represent words instead of the prediction on the next phoneme. Preliminary experimental results are given. 15 Refs.", keywords = "Speech - Recognition Neural Networks - Applications Automata Theory - Computational Linguistics Voice Recognition Recurrent Neural Networks Lexical Decoding", } @InProceedings{cnlp:gera, title = "Netclass--{A} fresh look at connectionist category formation.", authorkey = "GeraMH", author = "M. H. Gera", year = "1992", pages = "225--230", publisher = "IEEE", address = "Piscataway, NJ", ISBN = "0-7803-0164-1", booktitle = "International Joint Conference on Neural Networks - IJCNN-91-Seattle Part 2 (of 2) Seattle, Washington, 8--12 July 1991", abstract = "The author describes Netclass, a proposed neural solution to the problem ofcategory formation and representation. Superordinates in Netclass have a fundamentally disjunctive representation. They are also strongly related to the actual scenes in which they are grouped. It is shown how these features, along with Netclass' connectionist nature, give a better account of some of the more recent data on superordinate categories than that offered by existing categorization models. Disjunction turns out to be of use in representing components. It is suggested that this offers a solution to a problem inherent in neural net concept component representation. 15 Refs.", keywords = "systems science and cybernetics - cognitive systems neural networks computer vision artificial intelligence learning systems artificial neural networks symbolic processing artificial intelligence models machine vision psychology", } @InProceedings{cnlp:roques, title = "Strategies of unsupervised learning for a parallel parsing architecture.", authorkey = "RoquesM BerouleD", author = "M. Roques and D. Beroule", year = "1992", pages = "215--218", publisher = "IEEE, IEEE Service Center, Piscataway, NJ, USA.", ISBN = "0-7803-0164-1", booktitle = "International Joint Conference on Neural Networks - IJCNN-91-Seattle Part 2 (of 2) Seattle, Washington", abstract = "A study concerning the analysis of natural language is reported, involving a serial and parallel processing principle already applied to pattern recognition tasks: guided propagation. The network of processing units which supports this principle is connected in the course of processing, thanks to basic differentiation and generalization mechanisms. It is shown how these basic learning mechanisms can be adapted so as to serve two purposes: the automatic clustering of unexpected words and the extraction of lexical substructures from examples of sentences through different possible strategies. Compared with classical natural language parsers the network that results from the learning mechanisms presented here implements a deterministic, left-to-right, breadth-first parser. The system follows every path simultaneously, in synchrony with the input flow. It works bottom-up from the input data, with a top-down trend, as pathways are reinforced. At a given time, part of the network tends to work in a top-down mode, while the other part works bottom-up, depending selectively on the 'familiarity' of the stored items. 4 Refs.", keywords = "Learning Systems - Analysis Systems Science AND Cybernetics - Man Machine Systems Natural Language Processing Systems Neural Networks Computer Systems, Digital - Parallel Processing Machine Learning Parsing Artificial Neural Networks Natural Language", } @InProceedings{cnlp:mcmillan, title = "Learning explicit rules in a neural network.", authorkey = "McMillanC MozerMC SmolenskyP", author = "C. McMillan and M. C. Mozer and P Smolensky", year = "1992", pages = "83--88", publisher = "IEEE", address = "Piscataway, NJ", ISBN = "0-7803-0164-1", booktitle = "International Joint Conference on Neural Networks - IJCNN-91-Seattle Part 2 (of 2) Seattle, Washington, 8--12 July 1991", abstract = "The authors propose an architecture called RuleNet, which, based on knowledge of the task domain, allows for the extraction of symbolic condition-action rules from the connection strengths in a neural net. By exploiting constraints inherent in the domain of symbolic string to string mappings, RuleNet can learn to induce explicit, symbolic, condition-action rules from examples. These rules represent a powerful representational language with which to describe the inner workings of a network. In addition, they facilitate faster learning and generalize perfectly to further examples that follow the rules. This formal string manipulation task can be viewed as an abstraction of several interesting cognitive models in the connectionist literature, such as case role assignment or translating English text into phonetic symbols. 7 Refs.", keywords = "neural networks - analysis learning systems systems science and cybernetics - cognitive systems artificial neural networks machine learning cognitive models symbolic processing rule based systems", } @InProceedings{cnlp:gorse, title = "Learning sequential structure with recurrent p{RAM} nets.", authorkey = "GorseD TaylorJG", author = "D. Gorse and J. G. Taylor", year = "1992", pages = "37--42", publisher = "IEEE", address = "Piscataway, NJ", ISBN = "0-7803-0164-1", booktitle = "International Joint Conference on Neural Networks - IJCNN-91-Seattle Part 2 (of 2) Seattle, WA, 08--12 July 1991", abstract = "Networks of probabilistic RAMs (pRAMs) may be trained using both gradient descent and reinforcement training rules. These two approaches are applied to the problem of learning a simple grammar from exposure to a finite set of grammatically correct strings, and it is seen that the combination of nonlinearity and stochasticity in the pRAM output functions enables a recurrent network to learn the grammar quickly and accurately. Both gradient descent and reinforcement training can be used to train a pRAM net to recognize a simple regular grammar (dual parity) with a speed which greatly exceeds that which can be achieved with networks of more conventional processor. In particular, it was demonstrated that the use of the stochastic features of the pRAM in reinforcement training would lead to a very significant reduction in training time if the system were implemented in hardware. 7 Refs.", keywords = "Learning Systems - Analysis Data Storage, Digital - Random Access Neural Networks Automata Theory - Grammars Mathematical Techniques - Algorithms Artificial Neural Networks Learning Algorithms Probabilistic Ram Neural Hardware Grammar Learning Language Learning", } @InProceedings{cnlp:muthusamy, title = "A segment-based approach to automatic language identification.", authorkey = "MuthusamyYK ColeRA GopalakrishnanM", author = "Y. K. Muthusamy and R. A. Cole and M. Gopalakrishnan", year = "1991", volume = "1", pages = "353--356", publisher = "IEEE, IEEE Service Center, Piscataway, NJ, USA (IEEE cat n 91CH2977-7).", ISBN = "0736-7791 0-7803-003-3", booktitle = "Proceedings of the 1991 International Conference on Acoustics, Speech, and Signal Processing - Icassp 91 Toronto, Ont, Canada", abstract = "A segment-based approach to automatic language identification is discussed which is based on the idea that the acoustic structure of languages can be estimated by segmenting speech into broad phonetic categories. Automatic language identification can then be achieved by computing features that describe the phonetic and prosodic characteristics of the language, and using these feature measurements to train a classifier to distinguish between languages. As a first step in this approach, a multilanguage, neural-network-based segmentation and broad classification algorithm using seven broad phonetic categories has been built. The algorithm was trained and tested on separate sets of speakers of American English, Japanese, Mandarin Chinese, and Tamil. It currently performs with an accuracy of 82.3% on the utterances of the test set. 7 Refs.", keywords = "Speech - Recognition Database Systems Computer Programming - Algorithms Automatic Language Identification Phonetic Characteristics Prosodic Characteristics", } @Article{cnlp:hilberg, title = "The network of human languages and its technological implementation.", authorkey = "HilbergW", author = "W. Hilberg", journal = "Frequenz", year = "1991", volume = "45", number = "11--12", pages = "275--284", ISBN = "0016-1136", address = "Technischen Hochschule Darmstadt, Germany", abstract = "A new class of networks is presented, which was derived from text structure analysis. These networks exhibit remarkable differences with symmetrical networks (e.g. hypercubes) as well as with neural networks. Its structure is defined generally by statistical rules. Diameter and mean diameter are both rather small, and the number of connections per node is low, i.e. these networks have favourable properties for technical applications, especially in areas where application specific solutions are desired. The generation of natural languages might be a typical problem of this kind. A great variety of individual network structures is possible (comparable with the number of individual languages), including structures which are so regular, that a random origin can not be supposed. (author abstract) 9 Refs. In German.", keywords = "Neural Networks Computer Simulation - Applications Statistical Methods - Applications Human Language Network TEXT Structure Analysis Natural Language Generation", } @InProceedings{cnlp:cawley, title = "Application of neural networks to cognitive phonetic modelling.", authorkey = "CawleyGC GreenADP", author = "G. C. Cawley and A. D. P. Green", year = "1991", number = "349", pages = "280--284", publisher = "IEE", address = "Stevenage, UK", ISBN = "0537-9987", booktitle = "2nd International Conference on Artificial Neural Networks, Bournemouth, UK, November 1991", abstract = "A neural network is used to generate control parameters for a parallel formant speech synthesizer, corresponding to a sequence of allophonic tokens. Training is to be accomplished using formant data obtained from both natural and synthetic speech. It is intended that theories of cognitive phonetics, currently being developed in the Department of Language and Linguistics at the University of Essex, will be used in order to improve the modelling of coarticulation. (author abstract) 8 Refs.", keywords = "Neural Networks - Applications Speech - Synthesis Cognitive Phonetic Modelling", } @InProceedings{cnlp:ludermir, title = "Logical neural nets and distributed implementations of weighted regular languages.", authorkey = "LudermirTB", author = "T. B. Ludermir", year = "1991", number = "349", pages = "158--162", publisher = "IEE", addresss = "Stevenage, UK", ISBN = "0537-9987", booktitle = "2nd International Conference on Artificial Neural Networks, Bournemouth, UK, November 1991", abstract = "A logical neural network, Aleksander (1), is a finite state machine then it is only possible to recognise regular grammars with these networks. When extra memory is associated with the nodes of these networks, the computational power of the model is increased and now weighted regular grammars, Salomaa (14), can be recognised. Through a constructive method based on the complexity of the production rules of the grammar, a logical network can be built to recognise any weighted regular language. The network generated by the constructive method is a distributed implementation of the weighted regular language. (author abstract) 14 Refs.", keywords = "neural networks automata theory - formal languages logical neural networks", } @InProceedings{cnlp:mueller, title = "Connectionist natural language parsing with Brain{C}.", authorkey = "MuellerA ZellA", author = "A. Mueller and A. Zell", year = "1991", pages = "188--196", publisher = "Int Soc for Optical Engineering, Bellingham, WA, USA.", ISBN = "0277-786X 0-8194-0578-7", address = "Univ. Stuttgart, 1 Stuttgart 80, Federal Republic of Germany", booktitle = "Applications of Artificial Neural Networks II Orlando, Florida", abstract = "A close examination of pure neural parsers shows that they either could not guarantee the correctness of their derivations or had to hard-code seriality into the structure of the net. The authors therefore decided to use a hybrid architecture, consisting of a serial parsing algorithm and a trainable net. The system fulfills the following design goals: (1) parsing of sentences without length restriction, (2) soundness and completeness for any context-free language, and (3) learning the applicability of parsing rules with a neural network to increase the efficiency of the whole system. BrainC (backtracktacking and backpropagation in C) combines the well- known shift-reduce parsing technique with backtracking with a backpropagation network to learn and represent typical structures of the trained natural language grammars. The system has been implemented as a subsystem of the Rochester Connectionist Simulator (RCS) on SUN workstations and was tested with several grammars for English and German. The design of the system and then the results are discussed. 7 Refs.", keywords = "Natural Language Processing Systems - Analysis Learning Systems Neural Networks Automata Theory - Computational Linguistics Artificial Neural Networks Parsing BACK Propagation Learning", } @InProceedings{cnlp:figuero, title = "Information representation analysis in a neural network.", authorkey = "FigueroaNazunoJ PerezElizaldeG VargasMedinaE RaggiGonzalezMG", author = "J. FigueroaNazuno and G. PerezElizalde and E. VargasMedina and M. G. RaggiGonzalez", year = "1991", pages = "2241--2246", publisher = "IEEE, IEEE Service Center, Piscataway, NJ, USA (IEEE cat n 91CH3065-0).", ISBN = "0-7803-0227-3", booktitle = "1991 IEEE International Joint Conference on Neural Networks - IJCNN '91, Singapore, November 1991", abstract = "The authors study the mathematical behavior of the hidden layer of a generalized delta rule type neural network (GDR) by analyzing the weights and thresholds in the network, when it learned and didn't learn, in a typical situation in neurocomputation. The GDR was used in a C language program. There are three representation hypotheses: a) the local, which states that information encoding takes place in local parts of the network; b) the generalized, which states that information is located in extended areas in the network; and c) the global, which states that total behavior represents the information in the networks. Several intensive computations were carried out to analyze the neural network internal behavior in situations where it did and didn't learn. The information shows clearly that representation as a global behavior in the hidden layer is responsible for learning, and not local behavior situations. 15 Refs.", keywords = "Neural Networks - Theory Computer Programming Languages - C Multilayer Neural Networks Hidden Layer Generalized Delta RULE Neural Network Information Representation Analysis", } @Article{cnlp:wu92, title = "Neural network feature maps for Chinese phonemes.", authorkey = "WuP WarwickK KoskaM", author = "P. Wu and K. Warwick and M. Koska", journal = "Neurocomputing", year = "1992", volume = "4", number = "1--2", pages = "109--112", ISBN = "0925-2312", address = "Univ of Reading, UK", abstract = "It has been shown through a number of experiments that neural networks can be used for a phonetic typewriter. Algorithms can be looked on as producing self-organizing feature maps which correspond to phonemes. In the Chinese language the utterance of a Chinese character consists of a very simple string of Chinese phoneme. With this as a starting point, a neural network feature map for Chinese phonemes can be built up. In this paper, feature map structures for Chinese phonemes are discussed and tested. This research on a Chinese phonetic feature map is important both for Chinese speech recognition and for building a Chinese phonetic typewriter. (Author abstract)", keywords = "Neural Networks - Applications Speech - Recognition Computer Programming - Algorithms Chinese Phonemes Feature MAPS", } @InProceedings{cnlp:mueller91, title = "Natural language parsing in a hybrid connectionist-symbolic architecture.", authorkey = "MuellerA ZellA", author = "A. Mueller and A. Zell", year = "1991", number = "Pt2", pages = "875--881", publisher = "Int Soc for Optical Engineering, Bellingham, WA, USA.", ISBN = "0277-786X", address = "Universitat Stuttgart, Stuttgart, Germany", booktitle = "Applications of Artificial Intelligence IX Orlando, Florida", abstract = "We here present a hybrid symbolic connectionist parser, which was designed to fulfill the following goals: 1) parsing of sentences without length restriction, 2) soundness and completeness for any context-free grammar, and 3) learning the applicability of parsing rules with a neural network. Our hybrid architecture consists of a serial parsing algorithm and a trainable net. BrainC (Backtracking and Backpropagation in C) combines the well known shift-reduce parsing technique with backtracking with a backpropagation network to learn and represent the typical properties of the trained natural language grammars. The system has been implemented as a subsystem of the Rochester Connectionist Simulator (RCS) on SUN-Workstations and was tested with several grammars for English and German. We discuss how brainC reached its design goals and what results we observed. (Edited author abstract) 7 Refs.", keywords = "Natural Language Processing Systems - Computer Applications Computer Programming - Algorithms Computer Architecture Neural Networks Connectionist Parsers Backpropagation Networks", } @InProceedings{cnlp:yin, title = "An associative memory model of language.", authorkey = "YinHF TaiJW", author = "H. F. Yin and J. W. Tai", year = "1990", pages = "663--670", publisher = "IEEE, IEEE Service Center, Piscataway, NJ, USA", address = "Inst of Autom, Academia Sinica, Beijing, China", booktitle = "1990 International Joint Conference on Neural Networks - IJCNN 90 Part 3 (of 3) San Diego, California", abstract = "An associative-memory model of language based on a neural network is proposed. It is shown that a language with finite sentences can be stored in a neural net completely. The memorizing process is realized by a dynamical learning algorithm which is convergent. Thus, a neural net not only has the ability to memorize syntactic information but can also memorize the semantic information of a language. The model is similar to human memory in some respects. A set of English words and Chinese sentences has been tested, and the simulation results on recognizing a word in default of a few characters are given. 2 Refs.", keywords = "DATA Storage, Digital - Associative Neural Networks Automata Theory Character Recognition Semantics Syntax Chinese Sentences English Words", } @InProceedings{cnlp:peschl, title = "A cognitive model coming up to epistemological claims: Constructivist aspects to modeling cognition.", authorkey = "PeschlMF", author = "M. F. Peschl", year = "1990", pages = "657--662", publisher = "IEEE, IEEE Service Center, Piscataway, NJ, USA (IEEE cat n 90CH2879-5).", address = "Dept for Epistemol \& Cognitive Sci, Univ of Vienna, Wien, Austria", booktitle = "1990 International Joint Conference on Neural Networks - IJCNN 90 Part 3 (of 3) San Diego, California", abstract = "An alternative approach to modeling cognition is discussed which considersnot only computer science, PDP (parallel distributed processing), and neuroscience aspects, but also epistemological issues. The aim is the construction of an adequate cognitive model satisfying epistemological as well as neuroscience claims. This project considers a constructivist point of view, implying that one must rethink the notions of knowledge representation, language, communication, etc. The proposed cognitive model is directly coupled to its environment; that is, there is only physical interaction, and no symbolic instance exists in between. This cognitive system is provided with several sensors and effectors allowing structural coupling to the environment (i.e., moving around, perceiving, emitting signals, etc.). Computer graphics methods are being used to model the retina. This approach seems to be better than the approaches taken by, for example, orthodox AI, because one is trying to understand and model the phenomenon of cognition from a bottom-up point of view, grasping basic processes first and only then simulating language, communications, etc. 15 Refs.", keywords = "SYSTEMS SCIENCE AND CYBERNETICS - Cognitive Systems AUTOMATA THEORY NEURAL NETWORKS COMPUTER GRAPHICS ARTIFICIAL INTELLIGENCE COGNITIVE MODELS EPISTEMOLOGY NEUROSCIENCE", } @Article{cnlp:miller, title = "Representing and computing regular languages on massively parallel networks.", authorkey = "MillerMI RoysamB SmithKR OsullivanJA", author = "M. I. Miller and B. Roysam and K. R. Smith and J. A. O'Sullivan", journal = "IEEE Transactions on Neural Networks", year = "1991", volume = "2", number = "1", pages = "56--72", ISBN = "1045--9227", address = "Dept of Electr Eng, Washington Univ, St Louis, Missouri", abstract = "A general method is proposed for incorporating rule-based constraints corresponding to regular languages into stochastic inference problems, thereby allowing for a unified representation of stochastic and syntactic pattern constraints. The authors' approach establishes the formal connection of rules to Chomsky grammars and generalizes the original work of Shannon on the encoding of rule-based channel sequences to Markov chains of maximum entropy. This maximum entropy probabilistic view leads to Gibbs representations with potentials which have their number of minima growing at precisely the exponential rate that the language of deterministically constrained sequences grow. These representations are coupled to stochastic diffusion algorithms, which sample the language-constrained sequences by visiting the energy minima according to the underlying Gibbs probability law. The coupling to stochastic search methods yields the all-important practical result that fully parallel stochastic cellular automata can be derived to generate samples from the rule-based constraint sets. The production rules and neighborhood state structure of the language of sequences directly determine the necessary connection structures of the required parallel computing surface. Representations of this type have been mapped to the DAP-510 massively parallel processor consisting of 1024 mesh-connected bit-serial processing elements for performing automated segmentation of electron-micrograph images. 66 Refs.", keywords = "NEURAL NETWORKS AUTOMATA THEORY - Formal Languages Probability - Random Processes IMAGE PROCESSING COMPUTER PROGRAMMING - Algorithms MASSIVELY PARALLEL NETWORKS RULE BASED CONSTRAINTS STOCHASTIC INFERENCE PROBLEMS", } @Article{cnlp:stjohn, title = "Learning and applying contextual constraints in sentence comprehension.", authorkey = "StJohnMF McClellandJL", author = "M. F. StJohn and J. L. McClelland", journal = "Artificial Intelligence", year = "1990", volume = "46", number = "1--2", pages = "217--257", ISBN = "0004-3702", address = "Carnegie-Mellon Univ, Pittsburgh, Pennsylvania", abstract = "A parallel distributed processing model is described that learns to comprehend single clause sentences. Specifically, it assigns thermatic roles to sentence constituents, disambiguates ambiguous words, intantiates vague words, and elaborates implied roles. The sentences are pre-segmented into constituent phrases. Each constituent is processed in turn to update an evolving representation of the event described by the sentence. The model uses the information derived from from each constituent to revise its ongoing interpretation of the sentence and to anticipate additional constituents. The network learns to perform these tasks through practice on processing example sentence/event pairs. The learning procedure allows the model to take a statistical approach to solving the bootstrapping problem of learning the syntax and semantics of a language from the same data. The model performs very well on the corpus of sentences on which it was trained, and generalizes to sentences on which it was not trained, but learns slowly. (author abstract) 35 Refs.", keywords = "SPEECH - Processing NEURAL NETWORKS LEARNING SYSTEMS SPEECH UNDERSTANDING CONNECTIONIST NETWORKS", } @Article{cnlp:derthick, title = "Mundane reasoning by settling on a plausible model.", authorkey = "DerthickM", author = "M. Derthick", journal = "Artificial Intelligence", year = "1990", volume = "46", number = "1--2", pages = "107--157", ISBN = "0004-3702", address = "MCC, Austin, Texas", abstract = "Connectionist networks are well suited to everyday common sense reasoning.Their ability to simultaneously satisfy multiple soft constraints allows them to select from conflicting information in finding a plausible interpretation of a situation. However these networks are poor at reasoning using the standard semantics of classical logic, based on truth in all possible models. This article shows that using an alternate semantics, based on truth in a single most plausible model, there is an elegant mapping from theories expressed using the syntax of propositional logic onto connectionist networks. An extension of this mapping to allow for limited use of quantifiers suffices to build a network from knowledge bases expressed in a frame language similar to KL-ONE. Although finding optimal models of these theories is intractable, the networks admit a hill climbing search algorithm that can be turned to give satisfactory answers in familiar situations. The article concludes with an example of retrieval involving incomplete and inconsistent information. Although this example works well, much remains before realistic domains are feasible. (Author abstract) 60 Refs.", keywords = "NEURAL NETWORKS COMPUTER METATHEORY - Formal Logic CONNECTIONIST NETWORKS KNOWLEDGE BASED SYSTEMS PROPOSITIONAL LOGIC REASONING", } @InProceedings{cnlp:gorin90, title = "On adaptive acquisition of language.", authorkey = "GorinAL LevinsonSE MillerLG GertnerAN LjoljeA GoldmanER", author = "A. L. Gorin and S. E. Levinson and L. G. Miller and A. N. Gertner and A. Ljolje and E. R. Goldman", year = "1990", volume = "1", pages = "601--604", publisher = "IEEE", address = "Piscataway, NJ", ISBN = "0736-7791", booktitle = "Proceedings of the 1990 International Conference on Acoustics, Speech, and Signal Processing - ICASSP 90 Albuquerque, NM, April 1990", abstract = "A system that automatically acquires a language model for a particular task from semantic-level information is described. This is in contrast to systems with predefined vocabulary and syntax. The purpose of the system is to map spoken or typed input into a machine action. To accomplish this task a medium-grain neural network is used. An adaptive training procedure is introduced for estimating the connection weights. It has the advantages of rapid, single-pass and order-invariant learning. The resulting weights have information-theoretic significance and do not require gradient search techniques for their estimation. The system was experimentally evaluated on three text-based tasks: a three-class inward-call manager with an acquired vocabulary of over 1600 words, a 15-action subset of the DARPA Resource Manager with an acquired vocabulary of over 700 words, and discrimination between idiomatic phrases meaning yes or no. 14 Refs.", keywords = "SPEECH - Recognition CONTROL SYSTEMS, ADAPTIVE - Estimation LEARNING SYSTEMS Database SYSTEMS NEURAL NETWORKS DARPA RESOURCE MANAGER LANGUAGE MODELS", } @InProceedings{cnlp:jain90b, title = "Robust connectionist parsing of spoken language.", authorkey = "JainAN WaibelAH", author = "A. N. Jain and A. H. Waibel", year = "1990", volume = "1", pages = "593--596", publisher = "IEEE", address = "Piscataway, NJ", ISBN = "0736-7791", address = "Sch of Comput Sci, Carnegie Mellon Univ, Pittsburgh, Pennsylvania", booktitle = "Proceedings of the 1990 International Conference on Acoustics, Speech, and Signal Processing - ICASSP 90 Albuquerque, New Mexico, April 1990", abstract = "A modular, recurrent connectionist network architecture which learns to robustly perform incremental parsing of complex sentences is presented. From sequential input, one word at a time, the networks learn to do semantic role assignment, noun phrase attachment, and clause structure recognition for sentences with passive constructions and center embedded clauses. The networks make syntactic and semantic predictions at every point in time, and previous predictions are revised as expectations are affirmed or violated with the arrival of new information. The networks induce their own grammar rules for dynamically transforming an input sequence of words into a syntactic/semantic interpretation. These networks generalize and display tolerance to input which has been corrupted in ways common in spoken language. 7 Refs.", keywords = "speech - recognition learning systems control systems - robustness automata theory - grammars robust connectionist parsing", } @InProceedings{cnlp:kim, title = "A study on the recognition of the Korean monothongs using artificial neural net models.", authorkey = "KimK KimI HwangH", author = "K. Kim and I. Kim and H. Hwang", year = "1990", pages = "364--371", publisher = "IEEE Computer Society Press", address = "Los Alamitos, CA", booktitle = "Proceedings of the 5th Jerusalem Conference on Information Technology, Jerusalem, Israel, October 1990", abstract = "The implementation and comparison of various artificial neural network models for recognition of Korean monothongs are reported. The goal is to develop an intelligent speech-based man-machine computer interface. The neural networks used were the multilayer perceptron, the time-delay neural net, the self-organizing feature map, and the interactive and competitive model. These four models were compared with respect to recognition rate and learning speed under various conditions. Experiments taking context effects, the most important problem in recognizing phonemes from continuous speech, into consideration were also performed. The models showed 90%-96% recognition rate for a male speaker. A strategy for Korean speech recognition using artificial neural networks has been developed on the basis of these results. 18 Refs.", keywords = "NEURAL NETWORKS - Applications NATURAL LANGUAGE PROCESSING SYSTEMS COMPUTATIONAL MODELS", } @Article{cnlp:minami, title = "Large-vocabulary spoken word recognition using time-delay neural network, phoneme spotting and predictive {LR}-parsing.", authorkey = "MinamiY SawaiH MiyatakeM", author = "Y. Minami and H. Sawai and M. Miyatake", journal = "Systems and Computers in Japan", year = "1991", volume = "22", number = "1", pages = "99--108", ISBN = "0882-1666", address = "Keio Univ, Yokohama, Japan", abstract = "This paper proposes a large-vocabulary speech recognition system using a phoneme spotting method by a time-delay neural network (TDNN) and a predictive LR parser. This is the first attempt to recognize large vocabulary speech using neural networks. The prediction of phonemes in words is performed by a predictive LR parser. Time alignment between predicted phonemes by the predicted LR parser and phoneme spotting results via TDNN is realized using a DTW (dynamic time warping) method. Speaker-dependent recognition for a 5240-word vocabulary using 2620 test words uttered by a male announcer resulted in a rate of 92.6 percent for the top choices, rates of 97.6 and 99.1 percent for the second and fifth choices, respectively. (author abstract)", keywords = "SPEECH - Recognition NEURAL NETWORKS - Applications TIME-DELAY NEURAL NETWORKS (TDNN) DYNAMIC TIME WARPING (DTW) PHONEME RECOGNITION", } @InProceedings{cnlp:namatae, title = "Connectionist learning with high-order functional networks and its internal representation.", authorkey = "NamatameA", author = "A Namatame", year = "1989", pages = "542--547", publisher = "IEEE, IEEE Service Center, Piscataway, NJ, USA. Available from IEEE Service Cent, Piscataway, NJ, USA.", ISBN = "0-8186-1984-8", address = "Natl Defense Acad, Dep of Comput Sci, Yokosuka, Japan", booktitle = "IEEE International Workshop on Tools for Artificial Intelligence: Architectures, Languages and Algorithms Fairfax, Virginia", abstract = "A novel architecture for supervised neural network learning is proposed.The necessary conditions of the network architecture for learning the structures of continuous mappings are obtained. The novel network architecture comprises high-order functional networks with some high-order functional units as input units. It is shown that high-order functional networks trained with backpropagation can generalize and infer the highly nonlinear structures of the continuous mappings. The internal representation capability of the high-order functional networks is analyzed. Nonlinear mappings can be characterized by the features of their extrema and curvatures. It is shown that the combination of the high-order functional input units and the hidden units makes it possible to realize and learn a proper internal representation of the networks for extracting these features of the continuous mappings. On the basis of these internal representation capabilities, a methodology for determining the network architecture and parameters is proposed. 14 Refs.", keywords = "SYSTEMS SCIENCE AND CYBERNETICS - Learning Systems COMPUTER SYSTEMS, DIGITAL - Parallel Processing ARTIFICIAL INTELLIGENCE CONNECTIONIST LEARNING SUPERVISED NEURAL NETWORK LEARNING NONLINEAR MAPPINGS CONTINUOUS MAPPINGS", } @InProceedings{cnlp:lange89, title = "Phase-locking of artificial neural oscillators can perform dynamic role-binding and inferencing.", authorkey = "LangeTE VidalJJ DyerMG", author = "T. E. Lange and J. J. Vidal and M. G. Dyer", year = "1989", pages = "595", publisher = "IEEE", address = "Piscataway, NJ", booktitle = "IJCNN International Joint Conference on Neural Networks, Washington, DC, June 1989", abstract = "Summary form only given, as follows. Previously, the authors had described a localist spreading-activation model, ROBIN (role binding and inferencing network), which uses stable, uniquely-identifying activation patterns, called signatures, to represent the dynamic role-bindings critical for high-level natural language understanding tasks. A unique signature activation on a node represents a role binding, which can be propagated as activation across long paths of nodes to allow inferencing. The authors illustrate that metastable artificial neural oscillators can be used to implement signature activations. In this novel model, groups of relaxation oscillators with unique patterns of natural oscillation frequencies serve as signatures. Phase-locking of gated, interacting oscillators allows the signatures to be propagated dynamically across the network for inferencing. Paths of synchronized oscillators form a chain of role bindings representing the model's plan/goal analysis of its natural language input.", keywords = "systems science and cybernetics - neural nets phase locked loops artificial neural oscillators natural language understanding tasks relaxation oscillators role binding signature activations", } @InProceedings{cnlp:samad, title = "Hybrid distributed/local connectionist architectures.", authorkey = "SamadT", author = "T Samad", year = "1989", pages = "583", publisher = "IEEE, IEEE Service Center, Piscataway, NJ, USA. Available from IEEE Service Cent (cat n 89CH2765-6), Piscataway, NJ, USA.", address = "Honeywell, Golden Valley, Minnesota", booktitle = "IJCNN International Joint Conference on Neural Networks Washington, DC", abstract = "Summary form only given, as follows. A class of neural-network architectures is described that uses both distributed and local representation. The distributed representations are used for input and output, thereby enabling associative, noise-tolerant interaction with the environment. Internally, all representations are fully local. This simplifies weight assignment and makes the networks easy to configure for specific applications. These hybrid distributed/local architectures are especially useful for applications were structured information needs to be represented. Three such applications are briefly discussed: a scheme for knowledge representation, a connectionist rule-based system, and a knowledge-base browser.", keywords = "SYSTEMS SCIENCE AND CYBERNETICS - Neural Nets COMPUTER SYSTEMS, DIGITAL - Distributed ARTIFICIAL INTELLIGENCE DISTRIBUTED/LOCAL CONNECTIONIST ARCHITECTURES CONNECTIONIST RULE-BASED SYSTEM KNOWLEDGE-BASE BROWSER KNOWLEDGE REPRESENTATION", } @InProceedings{cnlp:lee, title = "Learning distributed representations of conceptual knowledge.", authorkey = "LeeG FlowersM DyerM", author = "G. Lee and M. Flowers and M. Dyer", year = "1989", pages = "582", publisher = "IEEE", addresss = "Piscataway, NJ", booktitle = "IJCNN International Joint Conference on Neural Networks Washington, DC, June 1989", abstract = "Summary form only given, as follows. The authors argue that distributed representations must satisfy five criteria in order to serve as an adequate foundation for constructing and manipulating conceptual knowledge. These criteria are: automaticity, portability, structure encoding, semantic microcontent, and convergence. In our approach, distributed representations of semantic relations (i.e., propositions) are formed by recirculating the hidden layer in recurrent parallel distributed processing networks. The authors' experiments show that the resulting distributed semantic representations (DSRs) satisfy all of the above criteria. They believe that DSRs can help supply an important building block in developing more complex connectionist architectures for higher level inferencing, as is required in natural language processing.", keywords = "SYSTEMS SCIENCE AND CYBERNETICS - Neural Nets COMPUTER SYSTEMS, DIGITAL - Distributed CONCEPTUAL KNOWLEDGE REPRESENTATION DISTRIBUTED REPRESENTATIONS DISTRIBUTED SEMANTIC REPRESENTATIONS NATURAL LANGUAGE PROCESSING CONNECTIONIST ARCHITECTURES Abstract ONLY", } @InProceedings{cnlp:miikkulainen89, title = "Modular neural network architecture for sequential paraphrasing of script-based stories.", authorkey = "MiikkulainenR DyerMG", author = "R. Miikkulainen and M. G. Dyer", year = "1989", pages = "49--56", publisher = "IEEE", address = "Piscataway, NJ", booktitle = "IJCNN International Joint Conference on Neural Networks, Washington, DC, June 1989", abstract = "Sequential recurrent neural networks have been applied to a fairly high-level cognitive task, i.e., paraphasing script-based stories. Using hierarchically organized modular subnetworks, which are trained separately and in parallel, the complexity of the task is reduced by effectively dividing it into subgoals. The system uses sequential natural language input and output and develops its own I/O representations for the words. The representations are stored in an external global lexicon and are adjusted in the course of training by all four subnetworks simultaneously, according to the FGREP-method. By concatenating a unique identification with the resulting representation, an arbitrary number of instances of the same word type can be created and used in the stories. The system is able to produce a fully expanded paraphrase of the story from only a few sentences, i.e., the unmentioned events are inferred. The word instances are correctly bound to their roles, and simple plausible inferences of the variable content of the story are made in the process. 14 Refs.", keywords = "SYSTEMS SCIENCE AND CYBERNETICS - Cognitive Systems DATA PROCESSING MODULAR NEURAL NETWORK SEQUENTIAL RECURRENT NEURAL NETWORKS SEQUENTIAL PARAPHRASING NATURAL LANGUAGE PROCESSING FGREP-METHOD HIGH LEVEL COGNITIVE TASK", } @InProceedings{cnlp:weber, title = "A Connectionist model of conceptual representation.", authorkey = "WeberSH", author = "S. H. Weber", year = "1989", pages = "477--483", publisher = "IEEE, IEEE Service Center, Piscataway, NJ, USA. Available from IEEE Service Cent (cat n 89CH2765-6), Piscataway, NJ, USA.", address = "Dep of Comput Sci, Univ of Rochester, Rochester, New York", booktitle = "IJCNN International Joint Conference on Neural Networks Washington, DC", abstract = "A description is given of a connectionist architecture for modeling conceptual representation. Connectionist structures for capturing property-value bindings, conceptual aspects, scalar properties, property abstraction and inheritance, and category error detection are described. With these structures, the system known as DIFICIL (Direct Inferences and Figurative Interpretation in a Connectionist Implementation of Language comprehension) is able to perform direct inferences, both immediate and mediated, and interpret both literal and figurative adjective-noun combinations. 9 Refs.", keywords = "SYSTEMS SCIENCE AND CYBERNETICS - Neural Nets COMPUTER ARCHITECTURE CONNECTIONIST MODEL CONCEPTUAL REPRESENTATION DIFICIL SYSTEM LANGUAGE COMPREHENSION", } @InProceedings{cnlp:cole, title = "Language identification with neural networks: {A} feasibility study.", authorkey = "ColeRA InouyeJWT MuthusamyYK GopalakrishnanM", author = "R. A. Cole and J. W. T. Inouye and Y. K. Muthusamy and M. Gopalakrishnan", year = "1989", pages = "525--529", publisher = "IEEE", address = "Piscataway, NJ", booktitle = "IEEE Pacific RIM Conference on Communications, Computers and Signal Processing, Victoria, Canada, June 1989", abstract = "The feasibility of an approach to automatic language identification that combines recent advances in computer speech recognition and artificial neural networks is discussed. It is shown that artificial neural networks can be used as pattern classifiers that use information about distributions of broad phonetic categories to identify languages. Using artificial languages that differ only by their distribution of stop consonants, feature vectors were extracted from varying amounts of speech from each language. These feature vectors were then used to train an artificial neural network using the back-propagation algorithm. Classification results for two different sets of artificial languages are presented. 14 Refs.", keywords = "speech - recognition systems science and cybernetics - neural nets database systems computer speech recognition artificial neural networks stop consonants artificial languages back-propagation algorithm drive reinforcement neuron", } @InProceedings{cnlp:barnden88, title = "Simulations of Conposit, a supra-connectionist architecture for commonsense reasoning.", authorkey = "BarndenJA", author = "J. A. Barnden", year = "1988", pages = "311--315", publisher = "IEEE", address = "Piscataway, NJ", booktitle = "Proceedings: The 2nd Symposium on the Frontiers of Massively Parallel Computations, Fairfax, Virginia, October 1988", abstract = "A computation architecture called 'Conposit' is outlined. Composit manipulates very-short-term complex symbolic data structures of types that are useful in high-level cognitive tasks such as commonsense reasoning, planning, and natural language understanding. Conposit's data structures are, essentially, temporary configurations of symbol occurrences in a two-dimensional array of registers. Each register is implementable as a neural subnetwork whose activation pattern realizes the symbol occurrence. The data structures are manipulated by condition-action rules that are realizable as further neural subnetworks attached to the array. In simulations, Conposit performs symbolic processing of types previously found difficult for connectionist/neural networks. A version of Conposit, simulated on the massively parallel processor, embodying core aspects of P. Johnson-Laird's mental model theory (1983) of human syllogistic reasoning is concentrated on. This version illustrates Conposit's power and flexibility, which arises from two unusual data-structure encoding techniques: relative-position encoding and pattern-similarity association. 12 Refs.", keywords = "COMPUTER SYSTEMS, DIGITAL - Parallel Processing COMPUTER ARCHITECTURE SYSTEMS SCIENCE AND CYBERNETICS - Cognitive Systems ARTIFICIAL INTELLIGENCE - Applications COGNITIVE MODELING COMMONSENSE REASONING KNOWLEDGE REPRESENTATION NEURAL NETWORKS", } @Article{cnlp:wheately, title = "Connectionist model of children's comprehension and production of simple English sentences.", authorkey = "WheatleyB", author = "B. Wheatley", journal = "Neural Networks", year = "1988", volume = "1", number = "1 SUPPL", pages = "321", address = "Univ of Wisconsin-Milwaukee, Milwaukee, Wisconsin", booktitle = "International Neural Network Society 1988 First Annual Meeting Boston, Massachussetts", abstract = "Parallel distributed processing (connectionist) systems provide a promising model for complex cognitive systems like natural language. By constructing a connectionist network and observing its behavior, it is possible to model and test new hypotheses about long-standing questions in natural language research. Studies of young children have found that production of simple English sentences may be superior to comprehension in the absence of nonlinguistic cues. A Boltzmann machine network was constructed, trained, and tested for comprehension and production. The network consists of two overlapping subnetworks, one representing pragmatic knowledge and one representing syntactic knowledge. The set of units representing the meaning of a sentence is shared between the two subnetworks. 2 Refs.", keywords = "SYSTEMS SCIENCE AND CYBERNETICS - Cognitive Systems AUTOMATA THEORY SPEECH SUMMARY ONLY CONNECTIONIST MODEL CHILDREN'S COMPREHENSION ENGLISH SENTENCES BOLTZMANN MACHINE NETWORK SYNTACTIC KNOWLEDGE", } @Article{cnlp:charny, title = "Noun compound understanding using neural networks.", authorkey = "CharnyM", author = "M. Charny", journal = "Neural Networks", year = "1988", volume = "1", number = "1 SUPPL", pages = "293", address = "MITRE Corp, McLean, Virginia", booktitle = "International Neural Network Society 1988 First Annual Meeting Boston, Massachussetts", abstract = "Noun compounds are groups of two or more nouns which bring together separate concepts to form a new or altered concept, e.g., tea rose garden owner. Their meaning cannot be discovered by syntactic parsing but must employ semantic processing of some type. Understanding these compounds is a task humans generally do well but machines do poorly by contrast. They are an important part of natural language understanding and are a good testbed for the development of cognitive systems. Our current work employs a neural network paradigm to approach the problem of 'understanding' these noun compounds. We use a fully connected multilayer perceptron system with backward error propagation learning rule. 3 Refs.", keywords = "SPEECH - Recognition SYSTEMS SCIENCE AND CYBERNETICS - Cognitive Systems AUTOMATA THEORY SUMMARY ONLY NOUN COMPOUNDS SEMANTIC PROCESSING NATURAL LANGUAGE UNDERSTANDING MULTILAYER PERCEPTRON NEURAL NETWORKS", } @Article{cnlp:wang88, title = "Three neural models which process temporal information.", authorkey = "WangD KingIK", author = "D. Wang and I. K. King", journal = "Neural Networks", year = "1988", volume = "1", number = "1 SUPPL", pages = "227", address = "USC Dep of Computer Science, Los Angeles, California", booktitle = "International Neural Network Society 1988 First Annual Meeting Boston, Massachussetts", abstract = "Using temporal summation mechanism of a single synapse which is modeled and implemented in SLONN simulation language (Wang \& Hsu, 1988), we propose three neural models that can deal with three different aspects of temporal information processing. The first model will separate multiple temporal patterns into an array of neurons. Each neuron N//i in the array passes the temporal patterns whose frequencies are greater than and equal to certain value f//i. With one-sided lateral inhibition mechanism in the second model, each neuron N//i will only pass the temporal patterns whose frequencies are closest to f//i. Therefore, N//i behaves exactly as a frequency filter. By combining the second model with short-term memory mechanism (Grossberg, 1976) and associative learning, the third model can be used to store and recall sequences of temporal patterns. We describe them in some detail. (Edited author abstract) 2 Refs.", keywords = "SYSTEMS SCIENCE AND CYBERNETICS - Neural Nets INFORMATION THEORY COMPUTER SIMULATION LANGUAGES SUMMARY ONLY TEMPORAL INFORMATION SYNAPSE MULTIPLE TEMPORAL PATTERNS FREQUENCY FILTER TEMPORAL SUMMATION", } @InProceedings{cnlp:pham, title = "Mosaic: {A} macro-connectionist organization system for artificial intelligence computation.", authorkey = "PhamKM DegouletP", author = "K. M. Pham and P. Degoulet", year = "1988", pages = "533--540", publisher = "IEEE, New York, NY, USA. Available from IEEE Service Cent (cat n 88CH2632-8) Piscataway, NJ, USA.", address = "Hopital Broussais, Paris, France", booktitle = "IEEE International Conference on Neural Networks San Diego, California", abstract = "The authors present MOSAIC, an expert-system generator which uses amacroconnectionist approach to implement a dynamic system of knowledge representation. Each fragment of knowledge, termed a neuronic, integrates both declarative and procedural knowledge representation. The inference mechanism is fully distributed over the knowledge base, which eliminates the separation between knowledge and the treatment process. The macroconnectionist approach is derived from neurobiological models in which the basic unit corresponds to a system of neurons and not to a single neuron. This organization is positioned at an intermediate level between cognitive psychology and elementary neurobiology. However, the definition of a neuronic is recursive and makes it possible to integrate into the same system entities situated at different levels of complexity. MOSAIC uses both artificial-intelligence concepts (e.g., explicit inference strategies) and connectionist concepts (e.g., inference mechanisms based on the propagation of activation of dynamic contexts). A prototype based on these concepts has been constructed and its first medical application is being tested. 22 Refs.", keywords = "ARTIFICIAL INTELLIGENCE - Expert Systems SYSTEMS SCIENCE AND CYBERNETICS - Cognitive Systems MOSAIC SYSTEM MACROCONNECTIONIST APPROACH KNOWLEDGE REPRESENTATION NEUROBIOLOGICAL MODELS COGNITIVE PSYCHOLOGY", } @InProceedings{cnlp:nolfi, title = "Learning to understand sentences in a connectionist network.", authorkey = "NolfiS Parisi", author = "S. Nolfi and ?. Parisi", year = "1988", pages = "215--219", publisher = "IEEE, New York, NY, USA. Available from IEEE Service Cent (cat n 88CH2632-8) Piscataway, NJ, USA.", address = "Fondazione Sigma Tau, Rome, Italy", booktitle = "IEEE International Conference on Neural Networks San Diego, California", abstract = "Language understanding implies understanding of subject, verb, and object, or who did what to whom. The authors describe this as establishing the correct 'assembly links' between pairs of words in the input sentence. Some experiments are reported in which a network with memory units learns to understand sentences by incorporating during learning various constrains which are activated by word order, morphology, and prepositional marking of nouns. 5 Refs.", keywords = "SPEECH - Intelligibility SYSTEMS SCIENCE AND CYBERNETICS - Learning Systems DATA STORAGE UNITS CONNECTIONIST NETWORK LANGUAGE UNDERSTANDING", } @InProceedings{cnlp:ricotti, title = "Learning of word stress in a sub-optimal second order back-propagation neural network.", authorkey = "RicottiLP RagazziniS MartinelliG", author = "L. P. Ricotti and S. Ragazzini and G. Martinelli", year = "1988", pages = "355--361", publisher = "IEEE, New York, NY, USA. Available from IEEE Service Cent (cat n 88CH2632-8) Piscataway, NJ, USA.", address = "Fondazione Ugo Bordoni, Rome, Italy", booktitle = "IEEE International Conference on Neural Networks San Diego, California", abstract = "The authors show an example of an efficient and easy solution, using a neural network, of a problem that cannot be easily solved with rules. This example regards the localization of primary word stress. The knowledge of the position of primary stress is very useful in text-to-speech synthesis of Italian, a language characterized by a very prominent word accent. In fact, the position of word stress is the basis for the automatic generation of the pattern of duration of the syllables and of the intonation of the whole phrase. The authors use a feedforward network with an error back-propagation learning, extending the method with the computation of the correction step based on the second derivative of the error function. This method has been used to speed up convergence without using a fixed learning rate and a momentum term. The authors obtain a steep decrease of the error at the expense of a limited increase of the computational cost. To effectively use the method, it is, however, necessary to control some conditions that must be satisfied to use the second derivative as an estimate of the correction step. 3 Refs.", keywords = "SYSTEMS SCIENCE AND CYBERNETICS - Learning Systems SPEECH - Synthesis BACK-PROPAGATION NEURAL NETWORK TEXT-TO-SPEECH SYNTHESIS LEARNING OF WORD STRESS XOR PROBLEM ERROR BACK-PROPAGATION LEARNING", } @InProceedings{cnlp:lendaris, title = "Conceptual graph knowledge systems as problem context for neural networks.", authorkey = "LendarisGG", author = "G. G. Lendaris", year = "1988", pages = "133--140", publisher = "IEEE", address = "Piscataway, NJ", booktitle = "IEEE International Conference on Neural Networks, San Diego, California, July 1988", abstract = "For a connectionist network to be able to learn to generalize well, there must be some correspondence between the structure/constraints of the net's architecture and those of the given problem space. Therefore, recourse to experiments with real-world problems will always be required in connectionist research. The author gives a outline of a problem area for which connectionist nets hold great promise: knowledge systems, where the knowledge is encoded/represented using conceptual graphs. Certain aspects of this problem context are already known, and these are probed for possible implementation by connectionist nets. The approach used is to present some basic properties of conceptual graphs, indicate operations important in their application, and point out those that might be candidates for implementation with neural nets. A special representation schema for conceptual graphs is used for their implementation by neural nets. 8 Refs.", keywords = "systems science and cybernetics - neural nets database systems artificial intelligence conceptual graph knowledge systems connectionist network problem context", } @InProceedings{cnlp:smolensky87, title = "Analysis of distributed representation of constituent structure in connectionist systems.", authorkey = "SmolenskyP", author = "P. Smolensky", year = "1987", pages = "47", publisher = "IEEE, New York, NY, USA. Available from IEEE Service Cent (Cat n 87CH2386-1), Piscataway, NJ, USA", address = "Univ of Colorado, Boulder, Colorado", booktitle = "1987 IEEE Conference on Neural Information Processing Systems - Natural and Synthetic. Denver, Colorado", abstract = "Summary form only given, as follows. A general analysis of theconnectionist representation of complex symbolic structures (e. g., strings, stacks, trees) is reported. A general scheme is formulated called the tensor product representation, which generalizes a number of the representational approaches found in existing connectionist systems. The tensor product representation can be proved to possess a number of the properties that are required of a computationally adequate connectionist representation.", keywords = "SYSTEMS SCIENCE AND CYBERNETICS - Neural Nets COMPUTER ARCHITECTURE - Mathematical Models CONNECTIONIST SYSTEMS CONSTITUENT STRUCTURE DISTRIBUTED REPRESENTATION", } @InProceedings{cnlp:walters87a, title = "{RESPONSE} {MAPPING} {FUNCTIONS}: {CLASSIFICATION} {AND} {ANALYSIS} {OF} {CONNECTIONIST} {REPRESENTATIONS}.", authorkey = "WaltersD", author = "D Walters", year = "1987", pages = "III/79--86", publisher = "SOS Printing, San Diego, CA, USA. Available from IEEE Service Cent (cat n 87TH0191-7), Piscataway, NJ, USA", address = "State Univ of New York, Buffalo, New York", booktitle = "IEEE First International Conference on Neural Networks. San Diego, California", abstract = "A theoretical framework is presented in which connectionist representations can be classified and analyzed. The theoretical classification of connectionist variable representations has three purposes: to express existing variable representations in unified terms; to enable representations to be formally analyzed in a uniform manner; and to suggest new variable representations. In the the course of developing the theoretical framework for variable representations, it is shown that the process of measuring perceptual features or variables determines the type of connectionist representation for those features. This concept is further developed in a response mapping function analysis of distributed encodings.", keywords = "MATHEMATICAL TECHNIQUES - Function Evaluation CODES, SYMBOLIC - Encoding SYSTEMS SCIENCE AND CYBERNETICS - Neural Nets MAPPING FUNCTIONS CONNECTIONIST REPRESENTATIONS Variable REPRESENTATIONS DISTRIBUTED ENCODINGS THEORETICAL CLASSIFICATION", } @InProceedings{cnlp:chun, title = "A massively parallel model of schema selection.", authorkey = "ChunHW MimoA", author = "H. W. Chun and A. Mimo", year = "1987", pages = "II/379--386", booktitle = "IEEE First International Conference on Neural Networks. San Diego, CA, June 1987", abstract = "A massively parallel model of schema selection called SAMPAN is presented. The SAMPAN system is a constraint-satisfaction network with nodes that perform simple pattern matching and input summation. It is motivated by recent success in connectionist schema representations and in natural language marker-passing systems. A pure connectionist representation lacks generality; new propositions cannot easily be represented. SAMPAN gets around this problem by using marker-passing to perform variable binding on generalized concepts. This combination of marker-passing with connectionist spreading activation provides a highly malleable and general representation. 10 refs.", keywords = "Systems Science and Cybernetics - Neural Nets Computer Systems, Digital - Parallel Processing Artificial Intelligence Massively Parallel Model Schema Selection Sampan System Constraint-Satisfaction Network Connectionist Networks", } @InProceedings{cnlp:geffner, title = "On the probabilistic semantics of connectionist networks.", authorkey = "GeffnerH PearlJ", author = "H. Geffner and J. Pearl", year = "1987", pages = "II/187--195", publisher = "SOS Printing", address = "Available from IEEE,Piscataway, NJ", booktitle = "IEEE First International Conference on Neural Networks. San Diego, CA, June 9187", abstract = "The goodness/energy paradigm has recently emerged as a useful framework for the construction and analysis of connectionist models. Its lack of a clear semantics, however, has made it unsuitable as an specification language for the declarative content of those models. A correspondence is established between connectionist networks and a well-known family of probabilistic networks, thus endowing connectionist models with a well-understood probabilistic semantics. It is shown how a natural extension of the energy formulation presented by J. J. Hopfield (1982) leads to models capable of expressing arbitrary probability distributions. 12 refs.", keywords = "systems science and cybernetics - neural nets artificial intelligence probabilistic semantics connectionist networks goodness/energy paradigm probabilistic networks", } @Article{cnlp:partridge, title = "{CONNECTIONIST} {NETWORKS} {QUA} {GRAPHS}.", authorkey = "PartridgeD", author = "D. Partridge", journal = "Computers \& Mathematics with Applications 86 V15 N4 1988, First N M Symp on Graph Theor Models in Comput Sci, Las Cruces, NM, USA, Apr", year = "1988", pages = "325--331", ISBN = "0097-4943", address = "New Mexico State Univ, Las Cruces, New Mexico", abstract = "Connections networks are used to represent knowledge in terms of'subsymbolic' nodes. The connectionist paradigm is seen as a promising new approach to the realization of intelligent systems, and one that may be particularly amenable to formal analysis. This paper introduces connectionism, points out some of the major problems and argues that a graph-theoretical approach to some of the recognized problems may prove fruitful. (Edited author abstract) 17 refs.", keywords = "ARTIFICIAL INTELLIGENCE MATHEMATICAL TECHNIQUES - Graph Theory KNOWLEDGE REPRESENTATION SEMANTIC NETWORKS CONNECTIONIST NETWORKS", } @InProceedings{cnlp:gigley, title = "{SENTENCE} {COMPREHENSION} {PROCESSING} - {A} {SERIAL}-{ORDER}, {TIME}-{SYNCHRONOUS} {PROCESS}.", authorkey = "GigleyHM", author = "H. M. Gigley", journal = "Digest of Papers - IEEE Computer Society International Conference 32nd", year = "1987", pages = "39--42", publisher = "IEEE, New York, NY, USA. Available from IEEE Service Cent (Cat n 87CH2409-1), Piscataway, NJ, USA", ISBN = "0-8186-0764-5", address = "Univ of New Hampshire, Durham, New Hampshire", booktitle = "Digest of Papers - COMPCON Spring 87: Thirty-Second IEEE Computer Society International Conference. San Francisco, California", abstract = "A description is given of HOPE, a neuro-based computational model ofsingle-sentence-comprehension processing. An AI model, it focuses on syntactically based disambiguation within a processing paradigm based on the types of ambiguity found in neural processing. It addresses performance issues reported in psycholinguistic studies of normal performance as well as neurolinguistic ones under lesion conditions. HOPE includes a facility for studying processing desynchronization as a possible cause of lesion performance behavior as well as a facility for studying the effects of knowledge dissolution on language performance. 16 refs.", keywords = "SPEECH - Recognition SYSTEMS SCIENCE AND CYBERNETICS BIOMEDICAL ENGINEERING - Neurophysiology PATTERN RECOGNITION ARTIFICIAL INTELLIGENCE - Expert Systems NATURAL LANGUAGE PROCESSING (NLP) PSYCHOLINGUISTICS NEUROLINGUISTICS HOPE MODEL SENTENCE COMPREHENSION", } @InProceedings{cnlp:wong, title = "{TOWARD} {A} {MASSIVELY} {PARALLEL} {SYSTEM} {FOR} {WORD} {RECOGNITION}.", authorkey = "WongMK ChunHW", author = "M. K. Wong and H. W. Chun", year = "1986", pages = "1967--1970", publisher = "IEEE, New York, NY, USA. Available from IEEE Service Cent (Cat n 86 CH2243-4), Piscataway, NJ, USA", ISBN = "0736-7791", address = "GTE Lab Inc, Waltham, Massachussetts", booktitle = "ICASSP 86 - Proceedings, IEEE-IECEJ-ASJ International Conference on Acoustics, Speech, and Signal Processing. Tokyo, Japan", abstract = "The authors describe a massively parallel system for word recognition.Based on the connectionist network model adopted from cognitive science and artificial intelligence, the system consists of a large number of simple neuronlike processing units, or nodes, which represent words, phonetic segments, or phonetic features. The computation consists of constant updating of activation levels of all nodes, resulting from the excitatory links and inhibitory links between the nodes. Input to the system consists of frame-by-frame scores of similarity to a set of phonetic segments necessary for distinguishing between words in the vocabulary. These similarity scores are combined into phonetic feature indexes for each frame of speech as input to the feature nodes in the network. A linguistic knowledge base is built into the nentwork, allowing both data-driven processing and top-down prediction to cooperate or compete in working toward the correct lexical hypothesis. 12 refs.", keywords = "COMPUTER SYSTEMS, DIGITAL - Parallel Processing SPEECH - Recognition ARTIFICIAL INTELLIGENCE INFORMATION SCIENCE - Language Translation and Linguistics SYSTEMS SCIENCE AND CYBERNETICS - Neural Nets MASSIVELY PARALLEL SYSTEM WORD RECOGNITION FRAME-BY-FRAME SIMILARITY SCORES PHONETIC SEGMENTS PHONETIC FEATURE INDEXES", } @InProceedings{cnlp:li, title = "Massively parallel network-based natural language parsing system.", authorkey = "LiT ChunHW", author = "T. Li and H. W. Chun", year = "1987", pages = "401--408", publisher = "IEEE", address = "Piscataway, NJ", ISBN = "0-8186-0780-7", booktitle = "Second International Conference on Computers and Applications. Beijing, China, June 1987", abstract = "A massively parallel network-based parsing scheme (MPNP) organizedaccording to context-free grammar rules is described. The scheme combines the advantages of connectionist networks with marker-passing algorithms. The simple spreading-activation algorithm in connectionist networks provide a basis for an expectation propagation model. The marker-passing algorithm permits this parsing scheme to produce all alternatives in parallel and generates all plausible parses as the output. The approach extends massively parallel parsing models to handle recursive definitions of grammar and sentences of arbitrary length. In addition, it eliminates irrelevant deductions without explicit heuristic knowledge by combining a bottom-up parsing algorithm with a top-down expectation propagation model. The approach has been used successfully to parse both English and Chinese sentences. 16 refs.", keywords = "artificial intelligence computer programming - algorithms computer systems, digital - parallel processing information science - language translation and linguistics natural language parsing system spreading-activation algorithms connectionist networks marker-passing algorithm expectation propagation model", } @InProceedings{cnlp:walters87b, title = "Representation Of Variables In Connectionist Networks.", authorkey = "WaltersDKW", author = "D. K. W. Walters", year = "1987", pages = "698--702", publisher = "IEEE, New York, NY, USA. Available from IEEE Service Cent (Cat n 87CH2465-3), Piscataway, NJ, USA", ISBN = "0-8186-0777-X", address = "Univ of Buffalo, New York", booktitle = "Proceedings - First International Conference on Computer Vision. London, UK", abstract = "There are various types of fine-grain parallel architectures which areproving useful for artificial intelligence research, ranging from neural networks, through connectionist networks, to the Connection Machine. The style of computing that is possible with all such connectionist networks is different from that possible with uniprocessor systems, or with parallel systems containing a small number of processors. Just as a different style of computation is possible in fine-grain parallel systems, the styles of representation of variables that are natural for such parallelism are different from the types of representation natural for serial, or coarse-grain parallel processing. However, there has been no general theoretical analysis of connectionist representations. A language is provided in which different parallel representations can be accurately described and the differences between representations clarifed. The theoretical framework for variable representations is also useful for suggesting new types of representations. A theoretical framework is developed which can aid in the formal analysis of fine-grain parallel representations. 14 refs.", keywords = "ARTIFICIAL INTELLIGENCE SYSTEMS SCIENCE AND CYBERNETICS - Neural Nets IMAGE PROCESSING - Theory CONNECTIONIST NETWORKS PARALLEL ARCHITECTURES CONNECTION MACHINE", } @InProceedings{cnlp:shnelle, title = "{CONCURRENT} {PARSING} {IN} Programmable {LOGIC} {ARRAY} ({PLA}-) {NETS} - {PROBLEMS} {AND} {PROPOSALS}.", authorkey = "SchnelleH", author = "H. Schnelle", year = "1984", pages = "150--153", publisher = "Assoc for Computational Linguistics, Menlo Park, CA, USA", address = "Ruhr-Univ Bochum, Bochum, West Germany", booktitle = "10th International Conference on Computational Linguistics, 22nd Annual Meeting of the Association for Computational Linguistics, Proceedings of Coling 84. Palo Alto, California", abstract = "The author provides a conceptual and practical introduction into theprinciples of wiring or constructing special machines for language processing tasks instead of programming a universal machine. Construction would, in principle, provide higher descriptive adequacy in computationally based linguistics. (Edited author abstract) 2 refs.", keywords = "AUTOMATA THEORY - Computational Linguistics SYSTEMS SCIENCE AND CYBERNETICS - Neural Nets SIGNAL PROCESSING LOGIC CIRCUITS PROCESS-ORIENTED LINGUISTICS VLSI DESIGN METHODS CELLULAR AUTOMATA OPERATIONAL PRIMITIVES TEST PREDICTION Programmable LOGIC ARRAY (PLA)", } @InProceedings{cnlp:reilly84, title = "{CONNECTIONIST} {MODEL} {OF} {SOME} {ASPECTS} {OF} {ANAPHOR} {RESOLUTION}.", authorkey = "ReillyRG", author = "R. G. Reilly", year = "1984", pages = "144--149", publisher = "Assoc for Computational Linguistics, Menlo Park, CA, USA", address = "St. Patrick's Coll, Dublin, Irel", booktitle = "10th International Conference on Computional Linguistics, 22nd Annual Meeting of the Association for Computational Linguistics, Proceedings of Coling 84. Palo Alto, California", abstract = "The author describes some recent developments in language processing involving computational models which more closely resemble the brain in both structure and function. These models employ a large number of interconnected parallel computational units which communicate via weighted levels of excitation and inhibition. A specific model is described which uses this approach to process some fragments of connected discourse. (Edited author abstract) 7 refs.", keywords = "AUTOMATA THEORY - Computational Linguistics SYSTEMS SCIENCE AND CYBERNETICS - Brain Models CONNECTIONIST MODELS ANAPHORA NETWORK PROGRAMMING LANGUAGE (NPL) COGNITIVE MODELING WORD SENSE LEVEL SCHEMA LEVEL", } @InProceedings{cnlp:birnbaum, title = "Lexical ambiguity as a touchstone for theories of language analysis.", authorkey = "BirnbaumL", author = "L. Birnbaum", year = "1985", pages = "815--820", ISBN = "0-934613-02-8", booktitle = "IJCAI 85: Proceedings of the Ninth International Joint Conference on Artificial Intelligence. Los Angeles, California, August 1985", abstract = "This paper assesses several broad approaches to language analysis with respect to the problem of lexical ambiguity. The impact of the problem on both syntactic and semantic analysis is discussed, and several common methods for disambiguation, including the use of selectional restrictions and scriptal lexicons, are analyzed. Their shortcomings illustrate the need for complex inference to resolve ambiguity, which forms one of the key functional arguments in favor of integrating language analysis with memory and inference. However, it has proven surprisingly difficult to realize such an integrated approach in practice. The difficulty is shown to stem primarily from the theories of memory and inferential processing utilized. The implications for recent approaches to language analysis based on connectionist mechanisms are explored. Finally, the requirements imposed by lexical disambiguation on theories of memory and inferential processing are discussed. (Edited author abstract) 32 refs.", keywords = "INFORMATION SCIENCE - Language Translation and Linguistics LEXICAL AMBIGUITY NATURAL LANGUAGE ANALYSIS DISAMBIGUATION MEMORY INFERENTIAL PROCESSING", } @InCollection{cnlp:cottrell, title = "Viewing parsing as word sense discrimination: a connectionist approach.", authorkey = "CottrellGW SmallSL", author = "G. W. Cottrell and S. L. Small", booktitle = "Computational Models of Natural Language Processing", year = "1984", pages = "91--119", publisher = "North-Holland", address = "Amsterdam", ISBN = "0-444-87598-0", abstract = "This paper advocates the interdisciplinary development of a computationaltheory of human language comprehension and proposes a collection of initial constraints from which to start on such an enterprise. In order to satisfy these constraints, our modeling effort employs an architecture significantly different from the typical computer and closer to that of the human brain. We use a particular spreading activation or active semantic network scheme, called connectionism, which entails a massive number of appropriately connected computing units that communicate through weighted levels of excitation and inhibition. While such an architecture does not solve any problems per se, we believe that a number of questions become easier to set forth and more straightforward to solve. This paper surveys a number of fundamental language comprehension issues from the new perspective, and presents some simulation results of a parsing model based on these considerations. (Author abstract) 54 refs.", keywords = "automata theory - computational linguistics systems science and cybernetics - artificial intelligence human language comprehension natural language processing (nlp)", } @Article{cnlp:sharkey94b, title = "Opening the black-box of connectionist nets - Some lessons from Cognitive Science", authorkey = "SharkeyNE SharkeyAJC JacksonSA", author = "N. E. Sharkey and A. J. C. Sharkey and S. A. Jackson", address = "Univ Sheffield,Dept Comp Sci,Sheffield S10 2TN,S Yorkshire,England", journal = "Computer Standards \& Interfaces", year = "1994", volume = "16", number = "3", pages = "279--293", abstract = "Connectionist networks are not simply opaque black boxes with useful and efficient computational properties. Rather they have important internal structure that must be harnessed if their full potential as a novel computing technology is to be realised. First, the importance and usefulness of opening the black box is discussed and research is reviewed on how internal representations have been studied and used in the Cognitive Science literature. In the second section, a simple method for the geometrical analysis of decision space is presented. This shows connectionist networks as transparent boxes in which their computational properties are clear. The paper finishes with an example of how decision space diagrams can be useful for investigating under what circumstances it is best to adapt old weights for use in novel tasks", keywords = "CONNECTIONIST REPRESENTATION, BLACK BOX COMPUTING, DECISION SPACE ANALYSIS, NEURAL NETWORKS, CLUSTER ANALYSIS REPRESENTATION", } @Article{cnlp:zeng, title = "Discrete Recurrent Neural Networks For Grammatical Inference", authorkey = "ZengZ GoodmanRM SmythP", author = "Z. Zeng and R. M. Goodman and P. Smyth", address = "Caltech,Dept Elect Engn,Pasadena,CA,91125 JET PROP Lab,PASADENA,CA,91109", journal = "IEEE Transactions on Neural Networks", year = "1994", volume = "5", number = "2", pages = "320--330", abstract = "We describe a novel neural architecture for learning deterministic context-free grammars, or equivalently, deterministic pushdown automata. The unique feature of the proposed network is that it forms stable state representations during learning-previous work has shown that conventional analog recurrent networks can be inherently unstable in that they cannot retain their state memory for long input strings. We have recently introduced the discrete recurrent network architecture for learning finite-state automata. Here we extend this model to include a discrete external stack with discrete symbols. A composite error function is described to handle the different situations encountered in learning. The pseudo-gradient learning method (introduced in previous work) is in turn extended for the minimization of these error functions. Empirical trials validating the effectiveness of the pseudo-gradient learning method are presented, for networks both with and without an external stack. Experimental results show that the new networks are successful in learning some simple pushdown automata, though overfitting and non-convergent learning can also occur. Once learned, the internal representation of the network is provably stable; i.e., it classifies unseen strings of arbitrary length with 100% accuracy.", keywords = "NEURAL NETWORKS, ROBOT PATH PLANNING, MULTIVARIATE FUNCTION APPROXIMATION", } @Article{cnlp:gusgen, title = "Connectionist Inference Systems", authorkey = "GusgenHW HolldoblerS", author = "H. W. Gusgen and S. Holldobler", address = "Gesell Math \& Datenverarbeitung Gmbh,Schloss Birlinghoven,W-5205 St Augustin 1,Germany Th Darmstadt,Fb Informat,Fg Intellekt,W-6100 Darmstadt,Germany", journal = "Lecture Notes In Artificial Intelligence", year = "1992", volume = "590", pages = "82--120", abstract = "This paper presents a survey of connectionist inference systems.", keywords = "NEURAL NETWORKS, REPRESENTATION, OPTIMIZATION, RELAXATION, ALGORITHM, HOPFIELD", } @Article{cnlp:wallace, title = "Knowledge Representation And Decision-Making - {A} Hybrid Approach", authorkey = "WallaceI GossS BluffK", author = "I. Wallace and S. Goss and K. Bluff", address = "Swinburne Inst Technol,Pob 218,Hawthorn,Vic 3122,Australia Aeronaut Res Lab,Fishermans Bend,Vic 3207,Australia", journal = "Lecture Notes In Artificial Intelligence", year = "1992", volume = "604", pages = "525--528", abstract = "Knowledge representation is a fundamental issue in the simulation ofhuman performance such as the behaviour of a pilot engaged in aircombat. Interaction between symbolic and subsymbolic networks is used in a cognitive architecture to represent the effects of emotional and motivational processes on decision making. This enables a shift from the conventional focus of combat simulation on optimal decision making to the modelling of authentic human decision making with its individual variability and imperfection.", keywords = "CONTEXT-DEPENDENT SPELLING CORRECTION, GRAMMAR CHECKING, NATURAL-LANGUAGE-PROCESSING MODELS, NEURAL NET CLASSIFIERS, N-GRAM ANALYSIS, OPTICAL CHARACTER RECOGNITION (OCR), SPELL CHECKING, SPELLING ERROR DETECTION, SPELLING ERROR PATTERNS, STATISTICAL-LANGUAGE MODELS, WORD RECOGNITION AND CORRECTION", } @Article{cnlp:machado, title = "The Combinatorial Neural Network - {A} Connectionist Model For Knowledge Based Systems", authorkey = "MachadoRJ DarochaAF", author = "R. J. Machado and A. F. Darocha", address = "Ibm Brazil,Ctr Rio Sci,Estr Canoa 3520,Br-22610 Rio Janeiro,Brazil Unicamp,Inst Biol,Br-13100 Campinas,Brazil", journal = "Lecture Notes In Computer Science", year = "1991", volume = "521", pages = "578--587", abstract = "This paper describes the Combinatorial Neural Model, a high order neural network suitable for classification tasks. The model is based on the fuzzy sets theory, neural sciences and expert knowledge analysis results. The model presents Interesting properties such as: modularity, explanation capacity, concomitant knowledge and data representation, high speed of training, incremental learning, generalization capacity, feature selection, processing of uncertain and Incomplete data, fault tolerance.", keywords = "NEURAL NETWORKS, LEARNING, CONNECTIONIST EXPERT SYSTEMS", } @Article{cnlp:narazaki, title = "A Connectionist Approach For Rule-Based Inference Using An Improved Relaxation Method", authorkey = "NarazakiH RalescuAL", author = "H. Narazaki and A. L. Ralescu", address = "Tokyo Inst Technol,Dept Syst Sci,4259 Nagatsuta,Midori Ku,Yokohama,Kanagawa 227,Japan Lab Int Fuzzy Engn,Naka Ku,Yokohama 231,Japan", journal = "Ieee Transactions On Neural Networks", year = "1992", volume = "3", number = "5", pages = "741--751", abstract = "This paper describes a connectionist mechanism for an inference problem alternative to the usual chaining method. Our inference problem is within the scope of propositional logic that contains no variables and with some enhanced knowledge representation facilities. Basically, our method is an application of mathematical programming where knowledge and data are transformed into constraint equations. In our network, the nodes represent propositions and constraint equations, and the violation of constraints is formulated as an energy function The inference is realized as a minimization process of the energy function using the relaxation method to search for a truth value distribution that achieves the optimum consistency with the given knowledge and data. We propose a modified relaxation method to improve the computational inefficiencies associated with the optimization process. In contrast to the conventional microscopic inference technique based on local and piecewise evaluations of the knowledge, our method is macroscopic in that the whole knowledge is taken into consideration simultaneously. This aspect is vital to the inference under uncertainty or conflicting knowledge. We demonstrate and analyze the behavior of our method through examples of deductive and abductive inference and of inference with unorganized knowledge.", } @Article{cnlp:wyard, title = "Grammar Recognition By a Single Layer Higher-Order Neural Net", authorkey = "WyardPJ NightingaleC", author = "P. J. Wyard and C. Nightingale", address = "BT Labs,Martlesham Heath,Ipswich 1P5 7RE,Suffolk,England", journal = "BT Technology Journal", year = "1992", volume = "10", number = "3", pages = "77--96", abstract = "A single layer higher order dynamic topology neural net called HODYNE is described, targeted at natural language processing problems. HODYNE has an input layer of nodes containing concatenations of words from the input string, and grows nodes during training which are dependent on the training data. The training procedure also has some novel features.Results are presented on the problem of grammaticality determination. The results are encouraging, with almost perfect performance on unseen test sets derived from simple context-free grammars, and promising performance on sentences taken from the LOB Corpus of British English. On the latter task, the performance of HODYNE is compared with that of a method based on the statistics of tag-pairs and with a multilayer perceptron.", } @Article{cnlp:hummel, title = "Dynamic Binding in a Neural Network For Shape-Recognition", authorkey = "HummelJE BiedermanI", author = "J. E. Hummel and I. Biederman", journal = "Psychological Review", year = "1992", volume = "99", number = "3", pages = "480--517", abstract = "Given a single view of an object, humans can readily recognize that object from other views that preserve the parts in the original view. Empirical evidence suggests that this capacity reflects the activation of a viewpoint-invariant structural description specifying the object's parts and the relations among them. This article presents a neural network that generates such a description. Structural description is made possible through a solution to the dynamic binding problem: Temporary conjunctions of attributes (parts and relations) are represented by synchronized oscillatory activity among independent units representing those attributes. Specifically, the model uses synchrony (a) to parse images into their constituent parts, (b) to bind together the attributes of a part, and (c) to bind the relations to the parts to which they apply. Because it conjoins independent units temporarily, dynamic binding allows tremendous economy of representation and permits the representation to reflect the attribute structure of the shapes represented.", } @Article{cnlp:giles, title = "Learning and Extracting Finite State Automata with 2nd-Order Recurrent Neural Networks", authorkey = "GilesCL MillerCB ChenD ChenHH SunGZ LeeYC", author = "C. L. Giles and C. B. Miller and D. Chen and H. H. Chen and G. Z. Sun and Y. C. Lee", address = "NEC RES INST,4 Independence Way,Princeton,NJ,08540 Univ Maryland,Inst Adv Comp Studies,Dept Phys \& Astron,College PK,MD,20742", journal = "Neural Computation", year = "1992", volume = "4", number = "3", pages = "393--405", abstract = "We show that a recurrent, second-order neural network using areal-time, forward training algorithm readily learns to infer small regular grammars from positive and negative string training samples. We present simulations that show the effect of initial conditions, training set size and order, and neural network architecture. All simulations were performed with random initial weight strengths and usually converge after approximately a hundred epochs of training. We discuss a quantization algorithm for dynamically extracting finite state automata during and after training. For a well-trained neural net, the extracted automata constitute an equivalence class of state machines that are reducible to the minimal machine of the inferred grammar. We then show through simulations that many of the neural net state machines are dynamically stable, that is, they correctly classify many long unseen strings. In addition, some of these extracted automata actually outperform the trained neural network for classification of unseen strings.", } @Article{cnlp:reilly92, title = "A Connectionist Technique For Online Parsing", authorkey = "ReillyRG", author = "R. G. Reilly", address = "St Patricks Coll,Educ Res Ctr,Dublin 9,Ireland", journal = "Network-Computation in Neural Systems", year = "1992", volume = "3", number = "1", pages = "37--45", abstract = "A technique is described that permits the on-line construction and dynamic modification of parse trees during the processing of sentence-like input. The approach is a combination of simple recurrent network (SRN) and recursive auto-associative memory (RAAM). The parsing technique involves teaching the SRN to build RAAM representations as it processes its input item by item. The approach is a potential component of a larger connectionist natural language processing system, and could also be used as a tool in the cognitive modelling of language understanding. Unfortunately, the modified SRN demonstrates a limited capacity for generalization.", keywords = "NEURAL NETWORKS, INTERNAL REPRESENTATION, STATE SPACE, ORTHOGONAL COMPLEMENT METHOD, FEEDFORWARD NETWORK SYNTHESIS, THRESHOLD LOGIC", } @Article{cnlp:sowa, title = "Conceptual Graphs As {A} Universal Knowledge Representation", authorkey = "SowaJF", author = "J. F. Sowa", address = "IBM CORP,SYST RES,500 COLUMBUS AVE,THORNWOOD,NY,10594", journal = "COMPUTERS \& MATHEMATICS WITH APPLICATIONS", year = "1992", volume = "23", number = "2--5", pages = "75--93", abstract = "Conceptual graphs are a knowledge representation language designed as a synthesis of several different traditions. First are the semantic networks, which have been used in machine translation and computational linguistics for over thirty years, Second are the logic-based techniques of unification, lambda calculus, and Peirce's existential graphs. Third is the linguistic research based on Tesniere's dependency graphs and various forms of case grammar and thematic relations. Fourth are the dataflow diagrams and Petri nets, which provide a computational mechanism for relating conceptual graphs to external procedures and databases. The result is a highly expressive system of logic with a direct mapping to and from natural languages. The lambda calculus supports the definitions for a taxonomic system and provides a general mechanism for restructuring knowledge bases. With the definitional mechanisms, conceptual graphs can be used as an intermediate stage between natural languages and the rules and frames of expert systems-an important feature for knowledge acquisition and for help and explanations. During the past five years, conceptual graphs have been applied to almost every aspect of AI, ranging from expert systems and natural language to computer vision and neural networks. This paper surveys conceptual graphs, their development from each of these traditions, and the applications based on them.", } @Article{cnlp:shastri92, title = "Structured Connectionist Models of Semantic Networks", authorkey = "ShastriL", author = "L. Shastri", address = "Univ Penn,Dept Comp \& Informat Sci,Philadelphia,PA,19104", journal = "Computers \& Mathematics With Applications", year = "1992", volume = "23", number = "2--5", pages = "293--328", abstract = "In this paper, we review some connectionist realizations of semantic networks. We focus on models that are generally referred to as structured connectionist models. ln addition to reviewing three specific models, we discuss the intimate relationship between semantic networks and connectionist models of knowledge representation. We point out the need for massively parallel realizations of semantic networks and show that structured connectionism offers an appropriate computational framework for doing so.", keywords = "Semantic Networks, Expert System, Knowledge Processing Neural Networks, Fast Learning Algorithms, Impact of Connectionist Systems", } @Article{cnlp:pinker, title = "Rules of Language", authorkey = "PinkerS", author = "S. Pinker", address = "MIT,Dept Brain \& Cognit Sci,Cambridge,MA,02139", journal = "Science", year = "1991", volume = "253", number = "5019", pages = "530--535", abstract = "Language and cognition have been explained as the products of a homogeneous associative memory structure or alternatively, of a set of genetically determined computational modules in which rules manipulate symbolic representations. Intensive study of one phenomenon of English grammar and how it is processed and acquired suggests that both theories are partly right. Regular verbs (walk-walked) are computed by a suffixation rule in a neural system for grammatical processing; irregular verbs (run-ran) are retrieved from an associative memory.", } @Article{cnlp:hanson90, title = "What Connectionist Models Learn - Learning and Representation In Connectionist Networks", authorkey = "HansonSJ BurrDJ", author = "S. J. Hanson and D. J. Burr", address = "Siemens Res CTR,Learning \& Knowledge Acquisit Grp,Princeton,NJ,08540 Bellcore,Artificial Intelligence \& Commun Res GRP,Morristown,NJ,07960", journal = "Behavioral and Brain Sciences", year = "1990", volume = "13", number = "3", pages = "471--488", } @Article{cnlp:hendler, title = "But What is the Substance of Connectionist Representation - Comments", authorkey = "HendlerJ", author = "J. Hendler", address = "Univ Maryland,Dept Comp Sci,College PK,MD,20742", journal = "Behavioral and Brain Sciences", year = "1990", volume = "13", number = "3", pages = "496--496", } @Article{cnlp:pinker88, authorkey = "PinkerS PrinceA", author = "S. Pinker and A. Prince", year = "1988", title = "On language and connectionism: Analysis of a parallel distributed processing model of language acquisition. Special Issue: Connectionism and symbol systems", journal = "Cognition", volume = "28", number = "1--2", pages = "73--193.", } @PhdThesis{cnlp:leung89, authorkey = "LeungHC", author = "H. C. Leung", school = "Massachusetts Institute of Technology", title = "The Use of Artificial Neural Networks for Phonetic Recognition", year = "1989", annote = "He used back prop nets to classify English phonemes cut from continuous speech. The phonemes are from Timit. The signal representation is an auditory model of Seneff consisting of synchrony envelopes and mean rate response. The representation is adjusted for pitch; the coefficients are normalized so that their sum, per frame, is 0. Input was averaged spectrum from vowel thirds, plus context specifying the phoneme on the left and right. Performance was 60 percent.", } @PhdThesis{cnlp:henseler93, authorkey = "HenselerJ", author = "Johan Henseler", title = "Connections, Neurons and Activation, The Organization of Representation in Artificial Neural Networks", school = "University of Limburg", year = "1993", address = "Maastricht, Netherlands", } @InProceedings{cnlp:kohonen88london, authorkey = "KohonenT", author = "Teuvo Kohonen", title = "Associative Memories and Representations of Knowledge as Internal States in Distributed Systems", booktitle = "European Seminar on Neural Computing, London, U.K., February 8-9", year = "1988", publisher = "IBC Technical Services Ltd.", } @InProceedings{cnlp:veelenturf92, authorkey = "VeelenturfLPJ", author = "L. P. J. Veelenturf", title = "Representation of Spoken Words in a Self-Organizing Neural Net", booktitle = "Twente Workshop on Language Technology 3: {C}onnectionism and Natural Language Processing, Enschede, The Netherlands, May 1992", year = "1992", editor = "Anton Nijholt Marc F. J. Drossaers", pages = "1--4", publisher = "Department of Computer Science, University of Twente", } @PhdThesis{cnlp:dorffner89, authorkey = "DorffnerG", author = "G. Dorffner", title = "A Sub-Symbolic Connectionist Model of Basic Language Functions", school = "Indiana University, Dept. of Computer Science", year = "1989", } @InCollection{cnlp:dorffner90, authorkey = "DorffnerG", author = "G. Dorffner", title = "A Radical View on Connectionist Language Modelling", booktitle = "Konnektionismus in Artificial Intelligence und Kognitionforschung", publisher = "Springer", address = "Berlin", pages = "217--220", year = "1990", } @InProceedings{cnlp:hare, authorkey = "HareM ElmanJ", author = "M. Hare and J. Elman", title = "A Connectionist Account of English Inflectional Morphology", booktitle = "Proceedings of the Fourteenth Annual Conference of the Cognitive Science Society, Bloomington, IN, July 1992", publisher = "Lawrence Erlbaum Associates", pages = "265--270", year = "1992", } @Article{cnlp:karen, authorkey = "KarenLFR", author = "L. F. R. Karen", title = "Identification of Topical Entities in Discourse:{A} Connectionist Approach to Attentional Mechanisms in Language", journal = "Connection Science", pages = "103--122", year = "1990", } @InProceedings{cnlp:lange92, authorkey = "LangeT WhartonC", author = "T. Lange and C. Wharton", title = "{REMIND}: Integrating Language Understanding and Episodic Memory Retrieval in a Connectionist Network", booktitle = "Proceedings of the Fourteenth Annual Conference of the Cognitive Science Society, Bloomington, IN, July 1992", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", pages = "576--581", year = "1992", } @InProceedings{cnlp:liberman, authorkey = "LibermanM", author = "M. Liberman", title = "Questions about Connectionist Models of Natural Language", booktitle = "Proceedings of the 24th Annual Meeeting of the Association for Computional Linguistics, New York, NY, July 1986", year = "1986", } @InCollection{cnlp:mozer93, authorkey = "MozerMC DasS", author = "M. C. Mozer and S. Das", title = "A Connectionist Symbol Manipulator that Discovers the Structure of Context-Free languages", editor = "S. J. et al. Hanson", booktitle = "Advances in Neural Information Processing Systems 5", publisher = "Morgan Kaufmann", pages = "863--870", year = "1993", } @InProceedings{cnlp:waltz86, authorkey = "WaltzD", author = "D. Waltz", title = "Connectionist Models for Natural Language Processing", booktitle = "Proceedings of the 24th Annual Meeting of the Assocation for Computational Linguistics", year = "1986", } @Book{cnlp:touretzky91, editor = "D. Touretzky", title = "Connectionist Approaches to Language Learning", publisher = "Kluwer Academic Publishers", series = "The Kluwer International Series in Engineering and Computer Science", year = "1991", ISBN = "0792392167", } @InProceedings{cnlp:andersoncw_87, authorkey = "AndersonCW", author = "C. W. Anderson", title = "Strategy Learning in Multilayer Connectionist Representations", booktitle = "Proceedings of the Fourth International Workshop on Machine Learning, Irvine, CA, June 1987", publisher = "Morgan Kaufmann", pages = "103--114", year = "1987", } @InProceedings{cnlp:elman89b, authorkey = "ElmanJL", author = "J. L. Elman", title = "Structured Representations and Connectionist Models", booktitle = "Proceedings of the Eleventh Annual Conference of the Cognitive Science Society", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", year = "1989", } @Article{cnlp:french92, authorkey = "FrenchRM", author = "R. M. French", title = "Semi-Distributed Representations and Catastrophic Forgetting in Connectionist Networks", journal = "Connection Science", volume = "4", number = "3--4", pages = "365--378", year = "1992", } @Article{cnlp:griffith, authorkey = "GriffithN ToddP", author = "N. Griffith and P. Todd", title = "Process and Representation in Connectionist Models of Musical Structure", journal = "Connection Science", volume = "6", number = "1--2", pages = "131--134", year = "1994", } @InProceedings{cnlp:mozer90, authorkey = "MozerMC", author = "M. C. Mozer", title = "Discovering Faithful 'Wickelfeature' Representations in a Connectionist Network", booktitle = "The Twelfth Annual Conference of the Cognitive Science Society, Cambridge, MA, July 1990", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", pages = "356--363", year = "1990", } @TechReport{cnlp:sharkey92b, authorkey = "SharkeyN", author = "N. Sharkey", title = "The Ghost in the Hybrid: {A} study of Uniquely Connectionist Representations", institution = "Computer Science Dept, University of Exeter", number = "228", year = "1992", } @InCollection{cnlp:sharkey92a, authorkey = "SharkeyNE SharkeyAJC", author = "N. E. Sharkey and A. J. C. Sharkey", title = "A modular design for connectionist parsing", editor = "F. J. Marc and A. N. Drossaers", booktitle = "Twente Workshop on Language Technology 3: Connectionism and Natural Language Processing, Enschede, The Netherlands, May 1992", pages = "87--96", publisher = "Department of Computer Science, University of Twente", year = "1992", } @InCollection{cnlp:das94, authorkey = "DasS CMM", author = "S. Das and Mozer M. C.", title = "A Hybrid Gradient-Descent/Clustering Technique for Finite State Machine Induction", editor = "J. D. Cowan and G. Tesauro and J. Alspector", booktitle = "Advances in Neural Information Processing Systems 6", publisher = "Morgan Kaufmann", pages = "19--26", year = "1994", } @Article{cnlp:stjohn89, authorkey = "McClellandJL JohnM TarabanR", author = "James L. McClelland and Mark {St. John} and Roman Taraban", title = "Sentence Comprehension: {A} Parallel Distributed Processing Approach", journal = "Language and Cognitive Processes", year = "1989", volume = "3/4", pages = "287--335", } @Article{cnlp:ward-aij, authorkey = "WardN", author = "Nigel Ward", title = "A Parallel Approach to Syntax for Generation", year = "1992", volume = "57", pages = "183--225", journal = "Artificial Intelligence", abstract = "To produce good utterances from nontrivial inputs a natural langauge generator should consider many words in parallel, which raises the question of how to handle syntax in a parallel generator. If a generator is incremental and centered on the task of word choice, then the role of syntax is merely to help evaluate the appropriateness of words. One way to do this is to represent syntactic knowledge as an inventory of ``syntactic constructions'' and to have many constructions active in parallel at run-time. If this is done then the syntactic form of utterances can be emergent, resulting from synergy among constructions, and there is no need to build up or manipulate representations of syntactic structure. This approach is implemented in FIG, an incremental generator based on spreading activation, in which syntactic knowledge is represented in the same network as world knowledge and lexical knowledge.", } @InCollection{cnlp:smythe, authorkey = "SmytheEJ", author = "E. J. Smythe", title = "Temporal Representations in a Connectionist Speech System", editor = "D. S. Touretzky", booktitle = "Advances in Neural Information Processing Systems", publisher = "Morgan Kaufmann", pages = "240--247", year = "1989", } @TechReport{cnlp:thornton93, authorkey = "ThorntonC", author = "C. Thornton", title = "How Much is Enough? {A} Connectionist Perspective on the Representation Debate", institution = "School of Cognitive and Computing Sciences", addresss = "University of Sussex, UK", number = "274", year = "1993", } @InCollection{cnlp:touretzky90a, authorkey = "TouretzkyDS ElvgrenG", author = "D. S. Touretzky and G. Elvgren", title = "Rule Representations in a Connectionist Chunker", editor = "D. S. Touretzky", booktitle = "Advances in Neural Information Processing Systems 2", publisher = "Morgan Kaufmann", pages = "431--438", year = "1990", } @Article{cnlp:dell88, authorkey = "DellGS", author = "G. S. Dell", title = "The retrieval of phonological forms in production: Test of predictions from a connectionist model", journal = "Journal of Memory and Language", year = "1988", volume = "27", pages = "124--142", } @InCollection{cnlp:tanenhaus87, authorkey = "TanenhausMK DellGS CarlsonG", author = "M. K. Tanenhaus and G. S. Dell and G. Carlson", address = "Cambridge, MA", booktitle = "Modularity in Knowledge Representation and Natural-Language Understanding", chapter = "4", editor = "J. L. Garfield", pages = "82--108", publisher = "The MIT Press", title = "Context Effects in Lexical Processing: {A} Connectionist Approach to Modularity", year = "1987", } @Article{cnlp:waltrous91, authorkey = "WaltrousRL", author = "R. L. Waltrous", journal = "Computer Speech and Language", number = "5", pages = "341--362", title = "Context-modulated vowel discrimination using connectionist networks", year = "1991", notes = "context modulation", } @InProceedings{cnlp:berg87, title = "A Parallel Natural Language Processing Architecture with Distributed Control", authorkey = "BergG", author = "G. Berg", year = "1987", pages = "487--495", booktitle = "Proceedings of the Ninth Annual Cognitive Science Society, Seattle, WA, July 1987", organization = "Cognitive Science Society", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", note = "Keywords: ``Autonomous'' semantic networks, spreading activation. Not connectionism, but could be interesting to connectionists.", } @PhdThesis{cnlp:pollack87b, title = "On Connectionist Models of Natural Language Processing", authorkey = "PollackJB", author = "Jordan B. Pollack", year = "1987", school = "Computer Science Department of the University of Illinois at Urbana-Champaign", note = "Also available at technical report MCCS-87-100 from Box 3CRL, Computing Research Laboratory, New Mexico State University, Las Cruces, NM 88003", } @InProceedings{cnlp:chun87a, title = "A model of schema selection using marker passing and connectionist spreading activation", authorkey = "ChunHW MimoA", author = "Hon Wai Chun and Alejandro Mimo", year = "1987", pages = "887--895", booktitle = "Proceedings of the Ninth Annual Cognitive Science Society Conference, Seattle, WA, June 1987", organization = "Cognitive Science Society", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", } @InProceedings{cnlp:charniak87, title = "A connectionist context-free parser which is not context-free, but then it is not really connectionist either", authorkey = "CharniakE SantosE", author = "E. Charniak and E. Santos", year = "1987", pages = "70--77", booktitle = "Proceedings of the Ninth Annual Cognitive Science Society Conference, Seattle, WA, July 1987", organization = "Cognitive Science Society", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", } @InProceedings{cnlp:dolan87, title = "Symbolic Schemata, Role Binding, and the Evolution of Structure in Connectionist Memories", authorkey = "DolanCP DyerMG", author = "Charles P. Dolan and Michael G. Dyer", pages = "II:287--298", year = "1987", booktitle = "Proceedings of the IEEE First International Conference on Neural Networks, San Diego, CA, June 1987", organization = "IEEE", } @InProceedings{cnlp:kalita87, title = "Generation of Simple Sentences in {E}nglish Using the Connectionist Model of Computation", authorkey = "KalitaJ ShastriL", author = "Jugal Kalita and Lokendra Shastri", year = "1987", pages = "555--565", booktitle = "Proceedings of the Ninth Annual Cognitive Science Society Conference, Seattle, WA, July 1987", organization = "Cognitive Science Society", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", } @TechReport{cnlp:lachter87, title = "The Relation Between Linguistic Structure and Theories of Language Learning --- {A} Constructive Critique of Some Connectionist Learning Models", authorkey = "LachterJ BeverT", author = "Joel Lachter and Thomas Bever", year = "1987", institution = "University of Rochester", address = "Rochester, NY 14627", number = "Cognitive Science Technical Report URCS 44", note = "Also published in Cognition, Volume 28, Numbers 1-2, March 1988", } @InProceedings{cnlp:miyata87, title = "Organization of Action Sequences in Motor Learning: {A} Connectionist Approach", authorkey = "MiyataY", author = "Yoshiro Miyata", year = "1987", booktitle = "Proceedings of the Ninth Annual Cognitive Science Society Conference, Seattle, WA, July 1987", organization = "Cognitive Science Society", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", anote = "Learning by chunking and action hierarchy building", } @InProceedings{cnlp:parisi87, title = "Connectionist Modelling of Syntactic Constraints on Sentence Processing", authorkey = "ParisiD NolfiS", author = "Domenico Parisi and Stefano Nolfi", pages = "II:507--512", year = "1987", booktitle = "Proceedings of the IEEE First International Conference on Neural Networks", organization = "IEEE", } @InProceedings{cnlp:norrod87, title = "Feedback-Induced Sequentiality in Neural Networks", authorkey = "NorrodFE OneillMC GatE", author = "Forrest E. Norrod and Michael C. {O'Neill} and Erann Gat", pages = "II:251--258", year = "1987", booktitle = "Proceedings of the IEEE First International Conference on Neural Networks", organization = "IEEE", } @Article{cnlp:rumelhart81, title = "An Interactive Activation Model of Context Effects in Letter Perception: {P}art 2. {T}he Contextual Enhancement Effect and Some Tests and Extensions of the Model", authorkey = "RumelhartDE McClellandJL", author = "David E. Rumelhart and James L. McClelland", journal = "Psychological Review", year = "1982", pages = "60--94", volume = "89", } @InProceedings{cnlp:smolensky87b, title = "On the Connectionist Reduction of Conscious Rule Interpretation", authorkey = "SmolenskyP", author = "Paul Smolensky", year = "1987", pages = "187--194", booktitle = "Proceedings of the Ninth Annual Cognitive Science Society Conference", organization = "Cognitive Science Society", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", } @InProceedings{cnlp:stjohn87, title = "Reconstructive Memory for Sentences: {A} {PDP} Approach", authorkey = "JohnMFS McClellandJL", author = "Mark F. St. John and James L. McClelland", year = "1987", editor = "Danny R. Moates and Richard Butrick", booktitle = "Proceedings of Inference OUIC 86", publisher = "Ohio University", } @InProceedings{cnlp:sutton85, title = "The Learning of World Models by Connectionist Networks", authorkey = "SuttonRS PinetteB", author = "Richard S. Sutton and Brian Pinette", year = "1985", booktitle = "Proceedings of the Seventh Annual Cognitive Science Society Conference", organization = "Cognitive Science Society", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", } @InProceedings{cnlp:touretzky86b, title = "Representing and Transforming Recursive Objects in a Neural Network, or ``{T}rees {\it do} Grow on {B}oltzmann Machines''", authorkey = "TouretzkyDS", author = "David S. Touretzky", year = "1986", booktitle = "Proceedings of the 1986 IEEE International Conference on Systems, Man, and Cybernetics", organization = "IEEE", note = "Atlanta, GA", } @InProceedings{cnlp:touretzky87b, title = "Representing Conceptual Structures in a Neural Network", authorkey = "TouretzkyDS", author = "David S. Touretzky", pages = "II:279--286", year = "1987", booktitle = "Proceedings of the IEEE First International Conference on Neural Networks", organization = "IEEE", } @InProceedings{cnlp:touretzky87c, title = "Symbol Structures in Connectionist Networks: Five Properties and Two Architectures", authorkey = "TouretzkyDS DerthickMA", author = "David S. Touretzky and Mark A. Derthick", year = "1987", booktitle = "Proceedings of the IEEE COMPCON", organization = "IEEE", } @InProceedings{cnlp:waltz87, title = "Connectionist Models: Not Just a Notational Variant, Not a Panacea", authorkey = "WaltzDL", author = "David L. Waltz", year = "1987", pages = "58--64", booktitle = "Proceedings of the Third Workshop on Theoretical Issues in Natural Language Processing(TINLAP-3)", publisher = "Association for Computational Linguistics", note = "Abstract: Connectionist models inherently include features and exhibit behaviors which are difficult to achieve with traditional logic-based models. Among the more important characteristics are: (1) the ability to compute nearest match rather than requiring unification or exact match; (2) learning; (3) fault tolerance through the integration of overlapping modules, each of which may be incomplete or fallible, and (4) the possibility of scaling up such systems by many orders of magnitude, to operate more rapidly or to handle much larger problems, or both. However, it is unlikely that connectionist models will be able to learn all of language from experience, because it is unlikely that a full cognitive system could be built via learning from an initially random network; any successful large-scale connectionist learning system will have to be to some degree ``genetically'' prewired.", } @InCollection{cnlp:hinton81, authorkey = "HintonGE", author = "G. E. Hinton", title = "Implementing Semantic Nets in Parallel Hardware", booktitle = "Parallel Models of Associative Memory", editor = "G. E. Hinton and J. A. Anderson", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", year = "1981", } @Article{cnlp:cottrell83, authorkey = "CottrellGW SmallSL", author = "G. W. Cottrell and S. L. Small", title = "A connectionist scheme for modeling word-sense disambiguation", journal = "Cognition and Brain Theory", volume = "6", pages = "89--120", year = "1983", } @MastersThesis{cnlp:palmer94b, authorkey = "PalmerDD", author = "David D. Palmer", title = "{SATZ} - An Adaptive Sentence Segmentation System", booktitle = "Master's Thesis, University of California at Berkeley, December 1994, available as UCB Technical Report CSD-94-846", school = "University of California at Berkeley", year = "1994", } @Article{cnlp:sun92, authorkey = "SunR", author = "R. Sun", title = "On Variable Binding in Connectionist Networks", journal = "Connection Science", volume = "4", number = "2", pages = "93--124", year = "1992", } @TechReport{cnlp:sun90, authorkey = "SunR", author = "R. Sun", title = "Integrating Rules and Connectionism for Robust Reasoning", institution = "Brandeis University, Computer Science Dept", number = "TRCS-90-154", year = "1990", } @Unpublished{cnlp:sun95a, authorkey = "SunR", author = "R. Sun", title = "Robust Reasoning: integrating Rule-based and Similarity-based Reasoning", note = "To be published in {\em Artificial Intelligence}", year = "1995", } @InProceedings{cnlp:stolcke89b, authorkey = "StolckeA", author = "A. Stolcke", title = "Processing Unification-Based Grammars in a Connectionist Framework", booktitle = "Proceedings of the Eleventh Annual Conference of the Cognitive Science Society", publishers = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", pages = "908--915", year = "1989", } @InProceedings{cnlp:weber95, authorkey = "WeberV WermterS", author = "V. Weber and S. Wermter", title = "Towards Understanding Spontaneous Dialog Utterances in a Hybrid Model", address = "Sheffield, UK", booktitle = "Proceedings of the Tenth Biennial Conference on Artificial Intelligence and Cognitive Science", pages = "(to appear)", year = "1995", } @InCollection{cnlp:kwasny92b, authorkey = "KwasnySC FaisalKA", author = "S. C. Kwasny and K. A. Faisal", title = "Symbolic parsing via Sub-Symbolic Rules", editor = "J. Dinsmore", booktitle = "Symbolic and Connectionist Paradigms: Closing the Gap", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", year = "1992", } @Article{cnlp:kwasny91b, authorkey = "KwasnySC FaisalKA", author = "S. C. Kwasny and K. A. Faisal", title = "Rule-based training of neural networks", journal = "Expert systems with Applications (Special Issue on Applying Artificial Neural Networks to Expert Systems)", volume = "2", number = "1", pages = "47--58", year = "1991", } @InProceedings{cnlp:kwasny93a, authorkey = "KwasnySC KalmanBL ChangN", author = "S. C. Kwasny and B. L. Kalman and N. Chang", title = "Distributed Patterns as Hierarchical Structures", booktitle = "Proceedings of the World Congress on Neural Networks, Portland, OR, July 1993", volume = "2", pages = "198--201", year = "1993", } @InProceedings{cnlp:kwasny93b, authorkey = "KwasnySC JohnsonS KalmanBL", author = "S. C. Kwasny and S. Johnson and B. L. Kalman", title = "An Adaptive Neural Network Parser.", editor = "C. Daglu and B. R. Burke and B. R. Fernandez and J. Ghosh", booktitle = "Artificial Neural Networks in Engineering (ANNIE '93), Intelligent Engineering Systems through Artificial Neural Networks, St Louis, MO, November 1993", volume = "3", publisher = "ASME Press", pages = "467--472", year = "1993", } @InProceedings{cnlp:kwasny92c, authorkey = "KwasnySC KalmanBL", author = "S. C. Kwasny and B. L. Kalman", title = "A recurrent deterministic parser", booktitle = "Proceedings of the 4th Midwest Artificial Intelligence and Cognitive Science Society Conference", year = "1992", } @TechReport{cnlp:kalman94, authorkey = "KalmanBL KwasnySC", author = "B. L. Kalman and S. C. Kwasny", title = "High Performance Training of Feedforward and Simple Recurrent Networks", institution = "St. Louis: Department of Computer Science, Washington University", number = "WUCS-94-29", year = "1994", } @Article{cnlp:kwasny95, authorkey = "KwasnySC KalmanBL", author = "S. C. Kwasny and B. L. Kalman", title = "Tail-Recursive Distributed Representations and Simple Recurrent Networks", journal = "Connection Science", volume = "7", number = "1", pages = "61--80", year = "1995", } @InCollection{cnlp:smolensky95, authorkey = "SmolenskyP", author = "P. Smolensky", title = "Reply: Constituent Structure and Explanation in an Integrated Connectionist/Symbolic Cognitive Architecture", booktitle = "Connectionism: Debates in Psychological Explanation", editor = "C. MacDonald and G. MacDonald", publisher = "Blackwell", address = "Oxford, UK", pages = "223--290", year = "1995", } @Book{cnlp:macdonald95, editor = "C. MacDonald and G. MacDonald", title = "Connectionism: Debates in Psychological Explanation", publisher = "Blackwell", address = "Oxford", year = "1995", } @Book{cnlp:sun95b, editor = "R. Sun and L. Bookman", title = "Computational Architectures Integrating Neural and Symbolic Processes", publisher = "Kluwer Academic", year = "1995", } @Article{cnlp:browne94, authorkey = "BrowneA PilkingtonJ", author = "A. Browne and J. Pilkington", title = "Unification Using a Distributed Representation", journal = "SIGART Bulletin", volume = "5", number = "1", pages = "33--42", year = "1994", } @Article{cnlp:frixone92, authorkey = "FrixoneM SpinelliG", author = "M. Frixone and G. Spinelli", title = "Connectionism and Functionalism: the importance of being a sybsymbolist", journal = "Journal of Experimental and Theoretical Artificial Intelligence", volume = "4", number = "1", pages = "3--17", year = "1992", } @InProceedings{cnlp:shultz94, authorkey = "ShultzTR BuckinghamD Oshima-TakaeY", author = "T. R. Shultz and D. Buckingham and Y. Oshima-Takae", editor = "S. J. Hanson and T. Petsihe and M. Kearns and R. K. Rivest", title = "A connectionist model of the learning of personal pronouns in English", booktitle = "Computational Learning Theory and Natural Learning Systems: Intersections between theory and experiment --- Proceedings of the Workshop, Berkeley, 1991, CA", volume = "2", pages = "347--362", publisher = "MIT Press", address = "Cambridge, MA", year = "1994", } @Book{cnlp:niklasson94, editor = "L. F. Niklasson and M. B. Bod\`{e}n", title = "Connectionism in a broad perspective -- selected papers from the Swedish Conference (University of Skovde, Sweden, 9--10 September 1992)", publisher = "Ellis Horwood", address = "Chichester, UK", year = "1994", } @PhdThesis{cnlp:park93, authorkey = "ParkJJ", author = "J. J. Park", title = "Neural Network Processing of Linguistic Symbols using a multi-level grading rule", school = "Florida State University", address = "Tallahassee", year = "1993", } @PhdThesis{cnlp:balogh94, authorkey = "BaloghIL", author = "Imre L. Balogh", title = "An analysis of a connectionist internal representation: {D}o {RAAM} networks produce truly distributed representations?", school = "New Mexico State University", address = "Las Cruces, NM", year = "1994", } @PhdThesis{cnlp:daugherty, authorkey = "DaughertyKG", author = "K. G. Daugherty", title = "Connectionist Inflectional Morphology: a network based account of the past tense", school = "University of Southern California", address = "Los Angeles", year = "1994", } @Article{cnlp:dorffner95, authorkey = "DorffnerG", author = "G. Dorffner", title = "On grounding language with neural networks", journal = "IEE Colloquium: Grounding Representations --- Integration of Sensory Information in NLP (AI and Neural Nets Digest)", number = "1995/103", pages = "471--473", publisher = "IEE, London, UK", year = "1995", } @InProceedings{cnlp:hester94, authorkey = "HesterKA BringmannMJ LonganD NiccolaiMJ NowackWJ", author = "K. A. Hester and M. J. Bringmann and D. Longan and M. J. Niccolai and W. J. Nowack", title = "The {P}redictive {RAAM}: a {RAAM} that can learn to distinguish sequences from a continuous input stream", booktitle = "WCNN'94: International Neural Networks Society, Annual Meeting, San Diego, June 1994", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", volume = "4", pages = "97--103", year = "1994", } @InProceedings{cnlp:kamimura91, authorkey = "KamimuraR", author = "R. Kamimura", title = "Recurrent Neural Network to acquire the grammatical competence", booktitle = "Proceedings of IJCNN'91, Seattle, July 8-12 1991", pages = "361--366", volume = "1", publisher = "IEEE", address = "Piscataway, NJ", year = "1991", } @InProceedings{cnlp:han91, authorkey = "HanYS YoonHS ChoiKS", author = "Youg S. Han and Hyun-Soo Yoon and Key Sun Choi", title = "Cognitive Knowledge Representation in Neural Nets", booktitle = "Proceedings of ACNN'91, Sydney, Australia, Feb 1991", year = "1991", pages = "293", } @InProceedings{cnlp:nwana92, authorkey = "NwanaHS", author = "H. S. Nwana", title = "Drawing from the shortcomings of Artificial Intelligence: some critical issues Connectionism must address.", booktitle = "Proceedings of EXPERSYS-92", address = "Houston Texas, USA and Paris, France", year = "1992", pages = "495--500", } @Article{cnlp:towell94, authorkey = "TowellGG ShavlikJW", author = "G. G. Towell and J. W. Shavlik", title = "Knowledge-based Artificial Neural Networks", journal = "Artificial Intelligence", volume = "70", number = "1--2", pages = "119--165", year = "1994", } @InProceedings{cnlp:miikkulainen94, authorkey = "MiikkulainenR", author = "R. Miikkulainen", title = "Integrated Connectionist Models: building {AI} systems on subsymbolic foundations", booktitle = "Proceedings of the Sixth International Conference on Tools with Artificial Intelligence, New Orleans, LA, November 1994", publisher = "IEEE Computer Society Press", address = "Los Alamitos, CA", pages = "231--232", year = "1994", } @InProceedings{cnlp:archambault94, authorkey = "ArchambaultD BassanoJC", author = "D. Archambault and J. C. Bassano", title = "A neural network for supervised learning of natural language grammar", booktitle = "Proceedings of the Sixth International Conference on Tools with Artificial Intelligence, New Orleans, LA, November, 1994", publisher = "IEEE Computer Society Press", address = "Los Alamitos, CA", pages = "267--273", year = "1994", } @Article{cnlp:perez94, authorkey = "PerezP SalinasD", author = "P. Perez and D. Salinas", title = "Storage of Structured Patterns in a neural network", journal = "Physical Review E", volume = "50", number = "5", pages = "4182--4186", year = "1994", } @InProceedings{cnlp:ultsch94, authorkey = "UltschA GuinaraesG WeberV", author = "A. Ultsch and G. Guinaraes and V. Weber", title = "Self-organising feature maps for logical unification", booktitle = "Moving Towards Expert Systems Globally in the 21st Century", address = "Lisbon/Estoril, Portugal", publisher = "Elmsford, NY, USA", pages = "1288--1294", year = "1994", } @Article{cnlp:idicula94, authorkey = "IdiculaSM", author = "S. M. Idicula", title = "A Connectionist Approach for Communicating with Databases in Natural Language", journal = "Adv. Model. Anal. B", volume = "31", number = "2", pages = "31--40", year = "1994", } @InProceedings{cnlp:craven93, authorkey = "CravenMW ShavlikJW", author = "M. W. Craven and J. W. Shavlik", title = "Extracting Symbolic Rules from Artificial Neural Networks", booktitle = "Proceedings of the Second International Workshop on Multistrategy Learning --- MSL-93, Harpers Ferry, WV, May 1993", pages = "207--217", year = "1993", } @InProceedings{cnlp:boyd93, authorkey = "BoydR DriscollJ SyvI", author = "R. Boyd and J. Driscoll and I. Syv", title = "Incorporating Semantics within a Connectionist Model and a Vector Processing Model", booktitle = "Second Text REtrieval Conference -- TREC-2, Gaithensburg, MD, August 1993", pages = "291--302", year = "1993", } @InProceedings{cnlp:cohen95, authorkey = "CohenME HudsonDL AndersonMF", author = "M. E. Cohen and D. L. Hudson and M. F. Anderson", title = "Integration of symbolic reasoning and neural network modelling for reasoning with uncertain information", booktitle = "Proceedings of the ISCA International Conference, Fourth Golden West Conference on Intelligent Systems", address = "San Francisco, CA, USA", publisher = "Raleigh, NC, USA: International Society for Computers and their Applications", pages = "179--183", year = "1995", } @InProceedings{cnlp:wermter94c, authorkey = "WermterS LochelM", author = "S. Wermter and M. Lochel", title = "Connectionist learning of flat syntactic analysis for speech/language systems", booktitle = "Proceedings of the International Conference on Artificial Neural Networks --- ICANN94", address = "Sorrento, Italy", publisher = "Springer Verlag, Berlin, Germany", pages = "941--944", volume = "2", year = "1994", } @InProceedings{cnlp:browne94a, authorkey = "BrowneA PilkingtonJ", author = "A. Browne and J. Pilkington", title = "Variable binding in a neural network using a distributed representation", booktitle = "Proceedings of the European Symposion on Artificial Neural Networks --- ESANN, Brussels, Belgium, April 1994", pages = "199--204", year = "1994", } @InProceedings{cnlp:soler94, authorkey = "SolerO HoeR", author = "O. Soler and R. Van Hoe", title = "Bar: {A} connectionist model of bi-lingual access representations", booktitle = "Proceedings of the International Conference on Artificial Neural Networks --- ICANN94", address = "Sorrento, Italy", publisher = "Springer Verlag, Berlin, Germany", pages = "263--267", volume = "1", year = "1994", } @InProceedings{cnlp:parfitt94, authorkey = "ParfittSH", author = "S. H. Parfitt", title = "An {ANN} model of anaphora: implications for nativism", booktitle = "Proceedings of the International Conference on Artificial Neural Networks --- ICANN94", address = "Sorrento, Italy", publisher = "Springer Verlag, Berlin, Germany", pages = "222--225", volume = "1", year = "1994", } @InProceedings{cnlp:degerlache94, authorkey = "GerlacheM SperdutiA StaritaA", author = "M. de Gerlache and A. Sperduti and A. Starita", title = "Encoding conceptual graphs by labeling {RAAM}", booktitle = "Proceedings of the International Conference on Artificial Neural Networks --- ICANN94, Sorrento, Italy, May 1994", publisher = "Springer Verlag", address = "Germany", pages = "272--275", volume = "1", year = "1994", } @InProceedings{cnlp:sanfeliu95, authorkey = "FeliuAS AlquezarR", author = "A. San Feliu and R. Alquezar", title = "Active grammatical inference: {A} new learning methodology", booktitle = "Shape, Structure and Pattern Recognition, Nahariya, Israel, October 1994", publisher = "World Scientific", address = "Singapore", pages = "191--200", year = "1995", } @InProceedings{cnlp:sugiyama95, authorkey = "SugiyamaS", author = "S. Sugiyama", title = "A neural network application to semantic and logic recognition", booktitle = "IEEE International Conference on Systems, Man and Cybernetics --- Intelligent Systems for the 21st Century", address = "Vancouver, BC, Canada", publisher = "IEEE, New York, NY:USA", pages = "3168--3173", volume = "4", year = "1995", } @InProceedings{cnlp:takane95, authorkey = "TakaneY Oshima-TakaneY ShultzTR", author = "Y. Takane and Y. Oshima-Takane and T. R. Shultz", title = "Network analyses: the case of first and second pronouns", booktitle = "IEEE International Conference on Systems, Man and Cybernetics --- Intelligent Systems for the 21st Century", address = "Vancouver, BC, Canada", publisher = "IEEE, New York, NY:USA", pages = "3594--3599", volume = "4", year = "1995", } @InProceedings{cnlp:ries95, authorkey = "RiesK BuoFD WangYY", author = "K. Ries and Finn Dag Buo and Ye Yi Wang", title = "Improved Language Modelling by Unsupervised Acquisition of Structure", booktitle = "Proceedings of the 1995 International Conference on Acoustics, Speech and Signal Processing", address = "Detroit, MI, USA", publisher = "IEEE, New York, NY, USA", pages = "193--196", volume = "1", year = "1995", } @InProceedings{cnlp:ding, authorkey = "DingL", author = "Liya Ding", title = "Neural Prolog: the concepts, construction and mechanism", booktitle = "IEEE International Conference on Systems, Man and Cybernetics --- Intelligent Systems for the 21st Century, Vancouver, Canada, October 1995", publisher = "IEEE", address = "New York, NY", pages = "3603--3608", volume = "4", year = "1995", } @InProceedings{cnlp:wu95, authorkey = "WuXY McTearM OjhaP", author = "Xin Yu Wu and M. McTear and P. Ojha", title = "A representational scheme for a hybrid natural language processing system", booktitle = "IEEE International Conference on Systems, Man and Cybernetics --- Intelligent Systems for the 21st Century", address = "Vancouver, BC, Canada", publisher = "IEEE, New York, NY:USA", pages = "3162--3167", volume = "4", year = "1995", } @InProceedings{cnlp:cooper94, authorkey = "CooperT", author = "T. Cooper", title = "A language for connectionist pattern-recognition", booktitle = "World Congress on neural networks-San Diego, International Neural Network Society Annual Meeting, San Deigo, CA, June 1994", volume = "3", pages = "156--161", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", ISBN = "0-80-581745-X", year = "1994", } @InProceedings{cnlp:miikkulainen94a, authorkey = "MiikkulainenR BijwaardD", author = "R. Miikkulainen and D. Bijwaard", title = "Parsing embedded clauses with distributed neural networks", booktitle = "Proceedings of the Twelfth National Conference on Artificial Intelligence, Seattle, WA, July 1994", volume = "1 and 2", pages = "858--864", publisher = "MIT Press", address = "Cambridge MA", ISBN = "0-26-261102-3", year = "1994", } @InProceedings{cnlp:buo94, authorkey = "BuoFD PolzinTS WaibelA", author = "F. D. Buo and T. S. Polzin and A. Waibel", title = "Learning complex output representations in connectionist parsing of spoken language", booktitle = "Proceedings of the International Conference on Acoustics, Speech and Signal Processing, April 1994", volume = "1", pages = "365--368", publisher = "IEEE", address = "Piscataway, NJ", ISBN = "0-78-031775-0", year = "1994", } @InProceedings{cnlp:daugherty94, authorkey = "DaughertyKG HareM", author = "K. G. Daugherty and M. Hare", title = "The English past tense --- is this the extent of language- learning by connectionist networks?", booktitle = "1994 IEEE International Conference on Neural Networks, Orlando, FL, Jun 1994", volume = "1--7", year = "1994", pages = "2303--2308", publisher = "IEEE", address = "Piscataway, NJ", ISBN = "0-78-031901-X", } @InProceedings{cnlp:boden95, authorkey = "BodenM NiklassonLF", author = "M. Bod\`{e}n and L. F. Niklasson", title = "Features of distributed representations for tree-structures --- {A} study of {RAAM}", booktitle = "Current trends in Connectionism --- Proceedings of the Swedish Conference on Connectionism, Skovde, Sweden, March, 1995", year = "1995", pages = "121--139", editor = "L. F. Niklasson and M. Bod\`{e}n", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", ISBN = "0-80-581997-5", } @InProceedings{cnlp:lyon95, authorkey = "LyonC DickersonB", author = "C. Lyon and B. Dickerson", title = "A fast partial parse of natural-language sentences using a connectionist method", booktitle = "Seventh Conference of the European Chapter of the Association for Computational Linguistics, Dublin, EIRE, March 1995", pages = "215--222", publisher = "Morgan Kaufmann", address = "Palo Alto, CA", ISBN = "1-55-860366-2", year = "1995", } @Article{cnlp:sun95c, authorkey = "SunR", author = "R. Sun", title = "Structuring knowledge in vague domains", journal = "IEEE Transactions on Knowledge and Data Engineering", year = "1995", volume = "7", number = "1", pages = "120--136", abstract = "In this paper, we propose a model for structuring knowledge in vague and continuous domains where similarity plays a role in coming up with plausible inferences. The model consists of two levels, one of which is an inference network with nodes representing concepts and links representing rules connecting concepts, and the other is a microfeature-based replica of the first level. Based on the interaction between the concept nodes and microfeature nodes in the model, inferences are facilitated and knowledge not explicitly encoded in a system can be deduced via mixed similarity matching and rule application. The model is able to take account of many important desiderata of plausible reasoning and produces sensible conclusions accordingly. Examples will be presented to illustrate the utility of the model in structuring knowledge to enable useful inferences to be carried out in several domains.", } @Article{cnlp:yeung95, authorkey = "YeungDS FongHS", author = "D. S. Yeung and H. S. Fong", title = "A knowledge matrix representation for a rule-mapped neural- network", journal = "Neurocomputing", year = "1995", volume = "7", number = "2", pages = "123--144", abstract = "Neural networks are noted for their learning and generalizing capabilities. However, their advancement and applicabilities are severely limited by the low comprehensibility of their internal knowledge. Previously, the authors have proposed a rule-mapped neural network model, by incorporating domain knowledge initially. This paper suggests a tool named 'knowledge matrix', that produces symbolic interpretations of the network's response to an input. These interpretations can be shown to enhance the reasoning power of the system. Moreover, the system knowledge can be refined explicably. The proposed approach is tested with a Chinese character structure recognition problem. This is an attempt to model a human learning process that is commonly observed in many situations. For example, a trainee may be given a set of provisional rules at the very beginning which is expected to be moderated in accordance with his forthcoming experience.", } @Article{cnlp:miikkulainen95, authorkey = "MiikkulainenR", author = "R. Miikkulainen", title = "Script-based inference and memory retrieval in subsymbolic story processing", journal = "Applied Intelligence", year = "1995", volume = "5", number = "2", pages = "137--163", abstract = "DISCERN is an integrated natural language processing system built entirely from distributed neural networks. It reads short narratives about stereotypical event sequences, stores them in episodic memory, generates fully expanded paraphrases of the narratives, and answers questions about them. Processing in DISCERN is based on hierarchically-organized backpropagation modules, communicating through a central lexicon of word representations. The lexicon is a double feature map system that transforms each orthographic word symbol into its semantic representation and vice versa. The episodic memory is a hierarchy of feature maps, where memories are stored ''one- shot'' at different locations. Several high-level phenomena emerge automatically from the special properties of distributed neural networks in this model. DISCERN learns to infer unmentioned events and unspecified role fillers, generates expectations and defaults, and exhibits plausible lexical access errors and memory interference behavior. Word semantics, memory organization, and appropriate script inferences are automatically extracted from examples. DISCERN shows that high- level natural language processing is feasible through integrated subsymbolic systems. Subsymbolic control of high- level behavior and representing and learning abstractions are the two main challenges in scaling up the approach to more open-ended tasks.", } @Article{cnlp:barnden95, authorkey = "BarndenJA", author = "J. A. Barnden", title = "High-level reasoning, computational challenges for connectionism, and the conposit solution", journal = "Applied Intelligence", year = "1995", volume = "5", number = "2", pages = "103--135", abstract = "Sophisticated symbol processing in connectionist systems can be supported by two primitive representational techniques called Relative-Position Encoding (RPE) and Pattern-Similarity Association (PSA), and a selection technique called Temporal- Winner-Take-All (TWTA). TWTA effects winner-take-all selection on the basis of fine signal-timing differences as opposed to activation-level differences. Both RPE and PSA are far the encoding of highly temporary associations between representations. RPE is based on the way activation patterns are positioned relative to each other within a network. Under PSA, two patterns are temporarily associated if they have (suitable) subpatterns that are (suitably) similar. The article shows how particular versions of the primitives are used to good effect in a system called Conposit/SYLL. This is a connectionist implementation of a slightly simplified version of a complex existing psychological theory, namely Johnson- Laird's account of syllogistic reasoning. The computational processes in this theory present a major implementational challenge to connectionism. The challenge lies in the mutability, multiplicity, and diversity of the working memory structures, and the elaborateness of the processing needed for them. Conposit/SYLL's techniques allow it to meet the challenge. The implementation of symbolic processing in Conposit/SYLL is an interesting application of connectionism partly because it significantly affects the design of the symbolic processing level itself. In particular, it encourages the use of associative as opposed to pointer-based data structures, and the use of random as opposed to ordered iteration over sets of data structures. In addition, the article discusses Conposit/SYLL's somewhat unusual variable- binding approach.", } @Article{cnlp:tani95, authorkey = "TaniJ FukumuraN", author = "J. Tani and N. Fukumura", title = "Embedding a grammatical description in deterministic chaos --- an experiment in recurrent neural learning", journal = "Biological Cybernetics", year = "1995", volume = "72", number = "4", pages = "365--370", abstract = "We consider the modeling process of a ''biological'' agent by combining the concepts of neuroinformatics and deterministic chaos. We assume that an agent observes a target process as a stochastic symbolic process, which is restricted by grammatical constraints. Our main hypothesis is that an agent would learn the target model by reconstructing pan equivalent quasi- stochastic process on its deterministic neural dynamics. We employed a recurrent neural network (RNN), which is regarded as an adjustable deterministic dynamical system. Then, we conducted an experiment to observe how the RNN learns to reconstruct the target process, represented by a stochastic finite state machine in the simulation. The result revealed the capability of the RNN to evolve, by means of learning, toward chaos, which is able to mimic a target's stochastic process. We precisely analyzed the evolutionary process as well as the internal representation of the neural dynamics obtained. This analysis enabled us to clarify an interesting mechanism of the self-organization of chaos by means of neural learning, and also showed how grammar can be embedded in the evolved deterministic chaos.", } @Article{cnlp:robinson95, authorkey = "RobinsonWS", author = "W. S. Robinson", title = "Brain symbols and computationalist explanation", journal = "Minds and machines", year = "1995", volume = "5", number = "1", pages = "25--44", abstract = "Computationalist theories of mind require brain symbols, that is, neural events that represent kinds or instances of kinds. Standard models of computation require multiple inscriptions of symbols with the same representational content. The satisfaction of two conditions makes it easy to see how this requirement is met in computers, but we have no reason to think that these conditions are satisfied in the brain. Thus, if we wish to give computationalist explanations of human cognition, without committing ourselves a priori to a strong and unsupported claim in neuroscience, we must first either explain how we can provide multiple brain symbols with the same content, or explain how we can abandon standard models of computation. It is argued that both of these alternatives require us to explain the execution of complex tasks that have a cognition-like structure. Circularity or regress are thus threatened, unless noncomputationalist principles can provide the required explanations. But in the latter case, we do not know that noncomputationalist principles might not bear most of the weight of explaining cognition. Four possible types of computationalist theory are discussed; none appears to provide a promising solution to the problem. Thus, despite known difficulties in noncomputationalist investigational we have every reason to pursue the search for noncomputationalist principles in cognitive theory.", } @Article{cnlp:lloyd95, authorkey = "LloydD", author = "D. Lloyd", title = "Consciousness - {A} connectionist manifesto", journal = "Minds and Machines", year = "1995", volume = "5", number = "2", pages = "161--185", abstract = "Connectionism and phenomenology can mutually inform and mutually constrain each other. In this manifesto I outline an approach to consciousness based on distinctions developed by connectionists. Two core identities are central to a connectionist theory of consciousness: conscious states of mind are identical to occurrent activation patterns of processing units; and the variable dispositional strengths on connections between units store latent and unconscious information. Within this broad framework, a connectionist model of consciousness succeeds according to the degree of correspondence between the content of human consciousness (the world as it is experienced) and the interpreted content of the network. Constitutive self- awareness and reflective self-awareness can be captured in a model through its ability to respond to self-reflexive information, identify self-referential categories, and process information in the absence of simultaneous input. The qualitative ''feel'' of sensation appears in a model as states of activation that are not fully discriminated by later processing. Connectionism also uniquely explains several specific features of experience. The most important of these is the superposition of information in consciousness - our ability to perceive more than meets the eye, and to apprehend complex categorical and temporal information in a single highly- cognized glance. This superposition in experience matches a superposition of representational content in distributed representations.", } @Article{cnlp:giles95, authorkey = "GilesCL ChenD SunGZ ChenHH LeeYC GoudreauMW", author = "C. L. Giles and D. Chen and G. Z. Sun and H. H. Chen and Y. C. Lee and M. W. Goudreau", title = "Constructive learning of recurrent neural networks --- Limitations of recurrent casade correlation and a simple solution", journal = "IEEE Transactions on Neural Networks", year = "1995", volume = "6", number = "4", pages = "829--836", abstract = "It is often difficult to predict the optimal neural network size for a particular application, Constructive or destructive methods that add or subtract neurons, layers, connections, etc, might offer a solution to this problem, We prove that one method, recurrent cascade correlation, due to its topology, has fundamental limitations in representation and thus in its learning capabilities, It cannot represent with monotone (i.e., sigmoid) and hard-threshold activation functions certain finite state automata, We give a ''preliminary'' approach on how to get ground these limitations by devising a simple constructive training method that adds neurons during training while still preserving the powerful fully-recurrent structure, We illustrate this approach by simulations which learn many examples of regular grammars that the recurrent cascade correlation method is unable to learn.", } @Article{cnlp:hillis95, authorkey = "HillisAE CaramazzaA", author = "A. E. Hillis and A. Caramazza", title = "Representation of grammatical categories of words in the brain", journal = "Journal of Cognitive Neuroscience", year = "1995", volume = "7", number = "3", pages = "396--407", abstract = "We report the performance of a patient who, as a consequence of left frontal and temporoparietal strokes, makes far more errors on nouns than on verbs in spoken output tasks, but makes far more errors on verbs than on nouns in written input tasks. This double dissociation within a single patient with respect to grammatical category provides evidence for the hypothesis that phonological and orthographic representations of nouns and verbs are processed by independent neural mechanisms. Furthermore, the opposite dissociation in the verbal output modality, an advantage for nouns over verbs in spoken tasks, by a different patient using the same stimuli has also been reported (Caramazza & Hillis, 1991). This double dissociation across patients on the same task indicates that results cannot be ascribed to ''greater difficulty'' with one type of stimulus, and provides further evidence for the view that grammatical category information is an important organizational principle of lexical knowledge in the brain.", } @Article{cnlp:vai95a, authorkey = "VaiMK XuZM", author = "M. K. Vai and Z. M. Xu", title = "Representing knowledge by neural networks for qualitative- analysis and reasoning", journal = "IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING", year = "1995", volume = "7", number = "5", pages = "683--690", abstract = "A systematic approach has been developed to construct neural networks for qualitative analysis and reasoning. These neural networks are used as specialized parallel distributed processors for solving constraint satisfaction problems. A typical application of such a neural network is to determine a reasonable change of a system after one or more of its variables are changed. A six-node neural network is developed to represent fundamental qualitative relations. A larger neural network can be constructed hierarchically for a system to be modeled by using six-node neural networks as building blocks. The complexity of the neural network building process is thus kept manageable. An example of developing a neural network reasoning model for a transistor equivalent circuit is demonstrated. The use of this neural network model in the equivalent circuit parameter extraction process is also described.", } @Article{cnlp:pinkas95, authorkey = "PinkasG", author = "G. Pinkas", title = "Reasoning, nonmonotonicity and learning in connectionist networks that capture propositional knowledge", journal = "Artificial Intelligence", year = "1995", volume = "77", number = "2", pages = "203--247", abstract = "The paper presents a connectionist framework that is capable of representing and learning propositional knowledge. An extended version of propositional calculus is developed and is demonstrated to be useful for nonmonotonic reasoning, dealing with conflicting beliefs and for coping with inconsistency generated by unreliable knowledge sources, Formulas of the extended calculus are proved to be equivalent in a very strong sense to symmetric networks (like Hopfield networks and Boltzmann machines), and efficient algorithms are given for translating back and forth between the two forms of knowledge representation, A fast learning procedure is presented that allows symmetric networks to learn representations of unknown logic formulas by looking at examples, A connectionist inference engine is then sketched whose knowledge is either compiled from a symbolic representation or learned inductively from training examples. Experiments with large scale randomly generated formulas suggest that the parallel local search that is executed by the networks is extremely fast on average. Finally, it is shown that the extended logic can be used as a high-level specification language for connectionist networks, into which several recent symbolic systems may be mapped, The paper demonstrates how a rigorous bridge can be constructed that ties together the (sometimes opposing) connectionist and symbolic approaches.", } @Article{cnlp:park95, authorkey = "ParkNS RobertsonD StenningK", author = "N. S. Park and D. Robertson and K. Stenning", title = "Extension of the temporal synchrony approach to dynamic variable binding in a connectionist inference system", journal = "Knowledge-Based Systems(UK)", year = "1995", volume = "8", number = "6", pages = "345--357", abstract = "The relationship between symbolism and connectionism has been one of the major issues in recent artificial intelligence research. An increasing number of researchers from each side have tried to adopt the desirable characteristics of the approach. A major open question in this field is the extent to which a connectionist architecture can accommodate basic concepts of symbolic inference, such as a dynamic variable binding mechanism and a rule and fact encoding mechanism involving nary predicates. One of the current leaders in this area is the connectionist rule-based system proposed by Shastri and Ajjanagadde. The paper demonstrates that the mechanism for variable binding which they advocate is fundamentally limited, and it shows how a reinterpretation of the primitive components and corresponding modifications of their system can extend the range of inference which can be supported. Our extension hinges on the basic structural modification of the network components and further modifications of the rule and fact encoding mechanism. These modifications allow the extended model to have more expressive power in dealing with symbolic knowledge as in the unification of terms across many groups of unifying arguments.", } @Article{cnlp:setiono95, authorkey = "SetionoR LiuH", author = "R. Setiono and H. Liu", title = "Symbolic representation of neural networks", journal = "Computer", year = "1996", volume = "29", number = "3", pages = "71", abstract = "The highly nonlinear nature of neural networks' input-to-output mapping makes it difficult to describe how they arrive at predictions. Thus, although their predictive accuracy is satisfactory for applications from finance to medicine, they have long been thought of as ''black boxes.'' The authors propose to understand a neural network via rules extracted from it. Their algorithm, NeuroRule, extracts rules from a standard feed-forward neural network, with network training and pruning via the simple, widely used back- propagation method. The extracted rules, a one-to-one mapping of the pruned network, are compact and comprehensible and do not involve weight values. The authors' experiments show that neural-network-based rules are as accurate and compact as decision-tree-based rules, which are widely regarded as explicit and understandable. Thus, using rules extracted by NeuroRule, neural networks become understandable and could lose their black-box reputation.", } @Article{cnlp:itsuki95, authorkey = "ItsukiN", author = "N. Itsuki", title = "Acquisition of internal representation by learning of identity- mapping using overload learning", journal = "Lecture Notes in Computer Science", year = "1995", volume = "930", pages = "971--978", abstract = "Acquisition of internal representation by neural networks that learn identity-mappings is one of important methods for feature-abstraction. It is, however, difficult to decide the number of hidden units of the networks. in this article, I show a way to apply the overload learning (OLL) technique to overcome this problem. Because OLL causes to reduce the number of effective dimensions of hidden patterns, networks can get reduced internal representation by learning. Moreover, I show that this technique is useful to get spatial representation from symbolic representation, and to integrate various types of information.", } @Article{cnlp:santos95, authorkey = "SantosJ OteroRP MiraJ", author = "J. Santos and R. P. Otero and J. Mira", title = "{NETTOOL} --- {A} hybrid connectionist symbolic development environment", journal = "Lecture Notes in Computer Science", year = "1995", volume = "930", pages = "658--665", abstract = "In this work we present an environment, NEITOOL, for the development of hybrid connectionist-symbolic systems, in which the connectionist representation is based on the same knowledge representation model as that of symbolic systems. The hybridation between the knowledge elements is local, the connectionist training algorthm is also localized in each element of the network so that the knowledge required for the learning process is a part of it. There is also the possibility of including capabilities for inferential level processing in the elements of the network.", } @Article{cnlp:markov91, authorkey = "MarkovZ", author = "Z. Markov", title = "A tool for building connectionist-like networks based on term unification", journal = "Lecture Notes in Artificial Intelligence", year = "1991", vol = "567", pages = "199--213", abstract = "The paper presents a network modeling tool called Net-Clause Language (NCL), integrating some connectionist-like and some symbolic processing features in a unified computational environment. Unlike the other connectionist symbol processing approaches, NCL represents and processes symbols not by implementing them as patterns of activity in a traditional neural network, rather it uses some connectionist ideas to organize symbolic computation in a more flexible way. The paper presents two examples (deductive inference and image processing) which show how the connectionist-like and symbolic features of NCL can benefit each from the other.", } @InProceedings{cnlp:kawada95, authorkey = "KawadaM", author = "M. Kawada", title = "A construction of neural-net based {AI} systems", booktitle = "Proceedings of the 1st IEEE Conference on the Engineering of Complex Systems, held jointly with 5th CSESAW, 3rd IEEE RTAW and 20thIFAC/IFIP WRTP, Fort Lauderdale, FL, November 1995", publisher = "IEEE Computing Society Press", address = "Los Alamitos, CA", pages = "424--427", year = "1995", } @InProceedings{cnlp:khosla95, authorkey = "KhoslaR DillonT", author = "R. Khosla and T. Dillon", title = "Knowledge Modeling in Integrated Symbolic-Connectionist System", booktitle = "IEEE International Conference on Systems, Man and Cybernetics --- Intelligent Systems for the 21st Century, Vancouver, Canada, October 1995", publisher = "IEEE", addresss = "New York, NY", pages = "3879--3883", volume = "5", year = "1995", } @Article{cnlp:fu95, authorkey = "FuL", author = "LiMin Fu", title = "Introduction to Knowledge Based Neural Networks", journal = "Knowledge Based Systems (UK)", volume = "8", number = "6", pages = "299--300", year = "1995", } @Article{cnlp:opitz95, authorkey = "OpitzDW ShavlikJW", author = "D. W. Opitz and J. W. Shavlik", title = "Dynamically adding symbolically meaningful nodes to knowledge based neural networks", journal = "Knowledge Based Systems (UK)", volume = "8", number = "6", pages = "301--311", year = "1995", } @Article{cnlp:andrews95, authorkey = "AndrewsR DiederichJ TickleAB", author = "R. Andrews and J. Diederich and A. B. Tickle", title = "Survey and critique of techniques for extracting rules from trained artificical neural networks", journal = "Knowledge Based Systems (UK)", volume = "8", number = "6", pages = "378--389", year = "1995", } @InProceedings{cnlp:michos95, authorkey = "MichosSE", author = "S. E. Michos", title = "A hybrid knowledge representation model in a natural language interface to {MSDOS}", booktitle = "Proceedings of the 7th Conference on Tools with Artificial Intelligence, Herndon, VA, November 1995", publisher = "IEEE Computing Society Press", address = "Los Alamitos, CA", pages = "480--483", year = "1995", } @InProceedings{cnlp:kopecz95, authorkey = "KopeczK", author = "K. Kopecz", title = "Unsupervised learning if sequences on maps with lateral connections", booktitle = "International Conference on Artificial Neural Networks ICANN '95", address = "Paris, France", volume = "1", pages = "431--436", year = "1995", } @InProceedings{cnlp:misic-flogel95, authorkey = "Misic-FlogelJ", author = "J. Misic-Flogel", title = "Temporal sequence processing: learning recognition and tracking using neural nets", booktitle = "International Conference on Artificial Neural Networks ICANN '95", address = "Paris, France", volume = "1", pages = "155--160", year = "1995", } @InProceedings{cnlp:blanchet95, authorkey = "BlanchetP AlexandreF", author = "P. Blanchet and F. Alexandre", title = "A hierarchical connectionist architecture for learning internal representations of behaviours", booktitle = "International Conference on Artificial Neural Networks ICANN '95", address = "Paris, France", volume = "2", pages = "183--188", year = "1995", } @InProceedings{cnlp:callan95, authorkey = "CallanRE Palmer-BrownD", author = "R. E. Callan and D. Palmer-Brown", title = "Analytical Derivation of a {(S)RAAM}", booktitle = "International Conference on Artificial Neural Networks ICANN '95", address = "Paris, France", volume = "2", pages = "485--490", year = "1995", } @Article{cnlp:vandervelde95, authorkey = "VeldeF", author = "F. van der Velde", title = "Symbol Manipulation with neural networks: production of a context free language using a modifiable working memory", journal = "Connection Science", volume = "7", number = "3--4", pages = "247--280", year = "1995", } @Article{cnlp:lin95, authorkey = "LinKHC Chen SooW", author = "K. H. C. Lin and Tung-bo Chen and Von Wun Soo", title = "Neural network learning and encoding of thematic role assignments in parsing of simple Chinese sentences", journal = "Journal of Information Sciences and Engineering", volume = "11", number = "1", pages = "109--126", year = "1995", } @InProceedings{cnlp:lyon95a, authorkey = "LyonC FrankR", author = "C. Lyon and R. Frank", title = "Neural network design for a natural language parser", booktitle = "International Conference on Artificial Neural Networks ICANN '95", address = "Paris, France", volume = "1", pages = "105--110", year = "1995", } @TechReport{cnlp:ho95a, authorkey = "HoEKS ChanLW", author = "Edward Kei Shiu Ho and Lai Wan Chan", title = "Confluent Preorder Parsing", institution = "Department of Computer Science and Engineering, The Chinese University of Hong Kong", address = "Shatin, N.T., Hong Kong", number = "CS-TR-95-03", year = "1995", } @TechReport{cnlp:ho95b, authorkey = "HoEKS ChanLW", author = "Edward Kei Shiu Ho and Lai Wan Chan", title = "Efficient Connectionist Representations of Syntactic Parse Trees for Grammatical Inference", institution = "Department of Computer Science and Engineering, The Chinese University of Hong Kong", address = "Shatin, N.T., Hong Kong", number = "CS-TR-95-15", year = "1995", } @InProceedings{cnlp:ho95c, authorkey = "HoEKS ChanLW", author = "Edward Kei Shui Ho and Lai Wan Chan", title = "Syntactic Parsing Using {RAAM}", booktitle = "Proceedings of World Congress on Neural Networks, Washington DC, July 1995", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", volume = "1", pages = "485--488", year = "1995", } @InProceedings{cnlp:ho95d, authorkey = "HoEKS ChanLW", author = "Edward Kei Shiu Ho and Lai Wan Chan", title = "Linearization + Confluent Inference = Efficient Connectionist Representations of Parse Trees for Grammatical Inference", booktitle = "Proceedings of International Symposium on Artificial Neural Networks, Hsinchu, Taiwan, Dec 18--20 1995", pages = "IS--24--IS--31", year = "1995", } @InProceedings{cnlp:ho96, authorkey = "HoEKS ChanLW", author = "Edward Kei Shiu Ho and Lai Wan Chan", title = "Confluent preorder parser as a finite state automata", booktitle = "Proceedings of International Conference on Artificial Neural Networks ICANN'96, Bocham, Germany, July 16--19 1996", publisher = "Springer-Verlag", address = "Berlin", pages = "899--904", year = "1996", } @TechReport{cnlp:blank91, authorkey = "BlankDS MeedenLA MarshallJB", author = "D. S. Blank and L. A. Meeden and J. B. Marshall", title = "Exploring the Symbolic/Subsymbolic Continuum: {A} Case Study of {RAAM}", institution = "Department of Computer Science, Indiana University", address = "Bloomington, Indiana", number = "TR 332", year = "1991", } @InCollection{cnlp:blank92, authorkey = "BlankDS MeedenLA MarshallJB", author = "D. S. Blank and L. A. Meeden and J. B. Marshall", title = "Exploring the Symbolic/Subsymbolic Continuum: {A} Case Study of {RAAM}", editor = "J Dinsmore", booktitle = "Symbolic and Connectionist Paradigms: Closing the Gap", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", pages = "113--148", year = "1992", } @InProceedings{cnlp:hadley92, authorkey = "HadleyRF", author = "R. F. Hadley", title = "Compositionality and Systematicity in Connectionist Language Learning", booktitle = "Proceedings of the Fourteenth Annual Conference of the Cognitive Science Society, Bloomington, IN, Jul 1992", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", pages = "659--664", year = "1992", } @InProceedings{cnlp:weber92, authorkey = "WeberV", author = "V. Weber", title = "Connectionist Unification with a distributed representation", booktitle = "Proceedings of the International Joint Conference on Neural Networks -- IJCNN '92, Beijing, China", publisher = "IEEE", address = "Piscataway, NJ", pages = "555--560", year = "1992", } @InCollection{cnlp:weber93a, authorkey = "WeberV", author = "V. Weber", title = "Connectionist Unifying Prolog", booktitle = "Artificial Neural Networks and Genetic Algorithms", editor = "Albrecht R. F. and Reeves C. R. and Steele N. C.", edition = "2nd", publisher = "Springer Verlag", pages = "213--220", year = "1993", } @InProceedings{cnlp:french94, authorkey = "FrenchRM", author = "R. M. French", title = "Dynamically constraining connectionist networks to produce distributed, orthogonal representations to reduce catastrophic interference", booktitle = "Proceedings of the Sixteenth Annual Conference of the Cognitive Science Society, Atlanta, GA, August 1994", publisher = "Lawrence Erlbaum Associates", address = "Hillsdale, NJ", editors = "Ram A. and Eiselt K.", pages = "335--340", year = "1994", } @Article{cnlp:hammerton98, authorkey = "HammertonJA", author = "J. A. Hammerton", title = "Holistic Computation: Reconstructing a muddled concept", journal = "Connection Science", volume = "10", number = "1", pages = "3--19", year = "1998", } @InCollection{cnlp:niklasson97, authorkey = "NiklassonL SharkeyNE", author = "L. Niklasson and N. E. Sharkey", title = "Systematicity and generalization in compositional connectionist representations", editor = "G. Dorffner", booktitle = "Neural Networks and a New Artificial Intelligence", publisher = "Thomson Computer Press", address = "London", pages = "217--232", year = "1997", } @Article{cnlp:narayanan96, authorkey = "NarayananS TaglianniGE PageEW", author = "S. Narayanan and G. E. Taglianni and E. W. Page", title = "Enhancing {MLP} networks using a distributed representation", journal = "IEEE Transactions on Systems, Man and Cybernetics B", volume = "26", number = "1", pages = "143--149", year = "1996", } @Book{cnlp:levy95, editor = "J. Levy and D. Baraktairis and J. Bullinaria and P. Cairn", title = "Connectionist Models of Memory and Language", publisher = "UCL Press", address = "London", year = "1995", ISBN = "1 85728 368 6", } @Article{cnlp:callan97, authorkey = "CallanRE Palmer-BrownD", author = "R. E. Callan and D. Palmer-Brown", title = "{(S)RAAM}: An Analytical Technique for Fast and Reliable Derivation of Connectionist Symbol Structure Representations", journal = "Connection Science", volume = "9", number = "2", pages = "139--159", year = "1997", } @InProceedings{cnlp:sperduti95, authorkey = "SperdutiA StaritaA CG", author = "A. Sperduti and A. Starita and Goller C.", title = "Learning Distributed Representations for the Classfication of Terms", booktitle = "Proceedings of the 14th International Joint Conference on Artificial Intelligence (IJCAI-95), Montreal, Canada, 20 -- 25 August 1995", editor = "C. S. Mellish", pages = "509--515", year = "1995", } @Article{cnlp:garcia-carpintero95, authorkey = "Garcia-CarpinteroM", author = "M. Garcia-Carpintero", title = "The philosophical import of connectionism, a critical notice of Andy Clark's Associative Engines", journal = "Mind and Language", volume = "10", number = "4", pages = "370--401", year = "1995", } @Article{cnlp:avner96, authorkey = "AvnerS", author = "S. Avner", title = "Extraction of comprehensive symbolic rules from a multi-layer perceptron", journal = "Engineering Applications of Artificial Intelligence", volume = "9", number = "2", pages = "137--143", year = "1996", } @Article{cnlp:kohonen90, authorkey = "KohonenT", author = "T. Kohonen", title = "The Self-Organising Map", journal = "Proceedings of the IEEE", volume = "78", number = "9", pages = "1464--1480", year = "1990", }