Prof Ingrid Zukerman "A Probabilistic Approach for Discourse Interpretation"

I will describe a probabilistic approach to argument interpretation that casts the selection of an interpretation as a model evaluation task. In selecting the best model, i.e., that with the highest posterior probability, the formalism balances conflicting factors: model complexity against data fit, and structure complexity against belief reasonableness. Our mechanism receives as input arguments entered through a web interface, and activates an anytime algorithm to produce candidate interpretations. These interpretations comprise inferences that connect the argument propositions, suppositions that make sense of the beliefs in the argument, and justifications that explain the inferences. Our user evaluations show that the interpretations produced by our system are acceptable, and that there is strong support for the postulated suppositions and justifications.