Kanji dictionaries, which need to present a large number of complex characters in an order that makes them accessible by users, traditionally use several indexing techniques that are particularly suited to the printed medium. Electronic dictionary technology provides the opportunity of both introducing new indexing techniques that are not feasible with printed dictionaries, and also allowing a wide range of index methods with each dictionary. It also allows dictionaries to be interfaced at the character level with documents and applications, thus removing much of the requirement for complex index methods. This paper surveys the traditional indexing methods, introduces some of the new indexing techniques that have become available with electronic kanji dictionaries, and reports on an analysis of the index usage patterns in a major WWW-based electronic kanji dictionary. This is believed to be the first such analysis conducted and reported.
Unlike languages written in alphabetic, syllabic or similar scripts, languages such as Japanese and Chinese, which are written using a large number of characters: hanzi in Chinese, kanji in Japanese, require two distinct sets of dictionaries. These are:
A typical learner of Japanese needs to have both forms of dictionary, and the process of "looking up" an unknown word often involves initially using the character dictionary to determine the pronunciation of one or more of the characters, then using that pronunciation as an index to a word dictionary, in a process that can be time-consuming and error-prone.
The advent of electronic dictionaries has had a considerable impact on Japanese dictionary usage:
This paper will concentrate on the issues associated with Japanese kanji dictionaries. Many of these also apply to Chinese.
2 Indexing a Kanji Dictionary
The general problem confronting the publication of kanji dictionaries is the large number of kanji in use and the absence of an intrinsic and recognized lexical order for kanji. In the post-war educational reforms in Japan, the number of kanji taught in schools was restricted to a basic 1,850, which has now been increased to 1,945. This set of kanji, along with a small set designated for use in personal names, accounts for all but a small proportion of kanji usage in modern Japanese. Many dictionaries and similar reference books compiled for students are based on this set (Sakade, 1961; Henshall, 1988; Halpern, 1999; etc.). The main computer character-set standard used in Japan, JIS X 0208 (JIS, 1997), which extends to less-common kanji including those used in places-names, has 6,355 kanji. This set is the basis for several kanji dictionaries (Nelson, 1997; Spahn & Hadamitzky, 1996), while larger sets of kanji are covered in many dictionaries, e.g. the Kodansha Daijiten (Ueda, 1963) has 14,900 kanji and the 13-volume Morohashi Daikanwajiten (Morohashi, 1989) has over 45,000 kanji.
In this paper, the term "primary index" has been used for the method of ordering the kanji entries, and "secondary index" has been used for cross-reference lists of kanji based on alternative ordering systems.
The major traditional indexing technique for kanji and hanzi dictionaries has been the radical system (bushu in Japanese), based on 214 elements plus about 150 variants. These elements are graphic components of the character that occur frequently enough to be used for indexing purposes. For example, the kanji 村 (mura: village) is identified by the 木 radical, and in a dictionary would be grouped with other kanji identified by that radical (札, 朷, 李, 松, etc.), with the grouped kanji ordered by the number of strokes in the remainder of the kanji. Radical systems have been used in Chinese character dictionaries for nearly 2,000 years, and the dominant 214-radical system was first used in the 康煕字典 (kangxi zidian) published in 1716.
Virtually all major kanji dictionaries published in Japan use the radical indexing method as the primary index, as do a number of dictionaries published elsewhere. Some dictionaries use modified or reduced sets of radicals. The technique is not simple to use, and some skill and practice is required in correctly identifying the radical and counting the residual strokes. The difficulty has been compounded by recent simplifications of the glyphs of the kanji, which in some cases have modified or eliminated the radical.
There are a number of other techniques used for indexing kanji in a dictionary:
A summary of the indices available in a selection of dictionaries and references is in Table 1. The "P" indicates the primary index and an "S" indicates a secondary index. (The original Nelson uses a slightly modified version of the traditional radical index, and the Spahn & Hadamitzky Kanji Dictionary uses a simplified 79-radical system.)
3 Electronic Kanji Dictionaries
As mentioned above, electronic kanji dictionaries have an increased number of indexing methods available, and in particular have navigational advantages over traditional paper dictionaries:
Figures 1 and 2 show the GUIs for the bushu and SKIP methods in the kanji dictionary module of the JWPce word-processor (Rosenthal, 2002).
Among the new indexing methods introduced with electronic kanji dictionaries are:
This file is inverted, enabling dictionary software to identify the kanji containing a particular selection of radicals. Figure 3 shows the multi-radical lookup GUI in JWPce, having identified the 怡 kanji from its components.
New indexing techniques, such as those described above, have to date been largely confined to non-commercial packages based on the author's KANJIDIC project files (Breen, 2004). The commercial electronic kanji dictionaries in Japan, which are typically based on published kanji dictionaries, usually only provide radical, reading and occasionally stroke-count indices.
4 Usage Patterns in an Electronic Kanji Dictionary
The availability of a large range of indexing techniques in an electronic kanji dictionary raises the question of how useful they actually are to users of such dictionaries, and which methods are preferred by users. With dictionaries provided as software packages, measurement of the usage of the differing indexing techniques would be limited to such things as surveys of users. To date no analysis appears to have been carried out on user preferences in indexing methods.
One form of electronic kanji dictionary which is amenable to the direct measurement of usage patterns is the kanji dictionary component of WWW-based Japanese dictionary, such as the WWWJDIC server (Breen, 2003) developed by the author. The WWWJDIC server provides over twenty indices to its database of over 13,000 kanji, including all the techniques mentioned earlier in this paper. The users are primarily students of Japanese and non-native speakers of Japanese.
The server code at the Monash University site was extended to provide detailed statistics of the accesses to the kanji dictionary module. Information was collected over a two-week period, during which time over 70,000 accesses to the kanji dictionary were made. Table 2 contains a breakdown of the accesses by index type. In the case of accesses using the multi-radical method, it is clear that users frequently have to make several selections of radicals to reach the correct kanji. From inspection of the raw statistics, it appeared that on average three accesses were made by each user of this method for each target kanji. Accordingly, the reported accesses for this method have been reduced to make a more meaningful comparison with the other methods. The "Direct" method involves access to the kanji in a word encountered in another dictionary function, whereas the "Cut-Paste" method refers to kanji transferred from another WWW page or application.
|Access Method||Access %|
(In 20.3% of the accesses recorded in Table 2, the user opted to make a follow-on search of one of the "word" dictionaries on the server using a kanji as a search key.)
These results are interesting for a number of reasons:
It is recognized that this survey of usage patterns reflects both the preferences of the particular set of users who have chosen to use it, and the biases introduced by the interface, which in the case of HTML forms is often not as easily used as, for example, a tailored GUI. It is, however, a strong indication of the sorts of indexing methods which are found to be useful by such a group. It is also worth noting that despite the clumsiness of the Multi-radical selection form, which has over 200 check-boxes, it is clearly among the most popular kanji index methods.
Kanji dictionaries have traditionally been published using indexing techniques developed for use in the printed medium. Electronic dictionary techniques provide the opportunity both to interface such dictionaries directly with text, and also to introduce new techniques more suited to the computer-human interface. Implementation of such techniques and the subsequent measurement of their usage in an environment where users can choose from a variety of indexing methods indicates a high level of acceptance of and preference for the new indexing techniques.
Breen, J.W. 1995. Building an Electronic Japanese-English Dictionary, JSAA Conference, Brisbane.
Breen, J.W. 2003. A WWW Japanese Dictionary, in "Language Teaching at the Crossroads", Monash Asia Institute, Monash University Press
Breen, J.W. 2004. KANJIDIC - Kanji Database Project, http://www.csse.monash.edu.au/‾jwb/kanjidic.html
Rosenthal, G. 2002. JWPce: Japanese Word Processor http://www.physics.ucla.edu/‾grosenth/jwpce.html
Halpern, J. 1990. New Japanese-English Character Dictionary, Kenkyusha/NTC.
Halpern, J. 1999. Kanji Learner's Dictionary, Kodansha
Henshall, K.G. 1988. A Guide to Remembering Japanese Characters, Tuttle.
Japanese Industrial Standards Committee. 1997. JIS X 0208-1997 7-bit and 8-bit Coded Kanji Sets for Information Interchange, Japanese Standards Association.
Morohashi, T. et al. (1989) Daikanwa Jiten, (Large Character Dictionary), Taishukan.
Nelson, A.N. 1974. The Modern Reader's Japanese-English Character Dictionary, (second revised edition), Tuttle.
Nelson, A.N. revised Haig, J.H. 1997. The New Nelson Japanese-English Character Dictionary, Tuttle.
Sakade, F. et al. 1961. A Guide to Reading & Writing Japanese, (second edition), Tuttle.
Shibano, K. et al. 1997, 2002. JIS Kanji Dictionary, (first and second editions), Japan Standards Association.
Spahn, M. & Hadamitzky, W. 1996. The Kanji Dictionary, Tuttle.
Spahn, M. & Hadamitzky, W. 1997. Kanji & Kana: A Handbook of the Japanese Writing System Tuttle.
Tanaka, Y. et al. 2000. Unicode Kanji Information Dictionary, Sanseido.
Ueda, K. et al. 1963. Daijiten, (Large Character Dictionary), Kodansha.