Plausible Expectations-Based Inference for Semantic Analysis

摘要

We present a project aiming at developing a semantic analyzer based on linguistic and world knowledge. The major sources of knowledge are a semantic dictionary, an ontology, a fact base, and a set of common sense axioms. We show the types of information stored in these resources, and demonstrate how they interact. As a case study, we take a simple but typical dialogue type, in which one of the interlocutors makes a proposal and the other one gives an indirect answer. The task of the analyzer is to interpret the answer as either the acceptance of the proposal or its rejection and, most importantly, to substantiate this interpretation. We show which knowledge is used and what reasoning should be performed in order to understand an indirect answer.