仮説生成機構を用いた未知語を含む文の解析

概要

論文の詳細を見る
The sentence analysis algorithm based on DCG has a weak point that the analysis itself fails if the analyzed sentence contains unknown words, namely the words not stored in the dictionary. But it is not realistic that every word in the analyzed sentence should be stored in the dictionary. We have tried to take advantage of the hypothesis generation mechanism for analyzing the sentence contalning unknown words. In this paper, we make use of a subset of ICR(b-ICR), which is an interpreter for explanation of given goals with hypothesis generation mechanism based on the SLD-resolution, as a meta-interpreter to DCG top-down algorithm, so that the analysis should not fail even if the analyzed sentence contains unknown words. The b-ICR checks up each of failing subgoal atom in SLD-resolution and if the atom can be matched to the specified pattern(permitted pattern), b-ICR regards it as hypothesis atom and continues the interpretation without failing. If the syntactic and semantic attributes of words is designed to permitted pattern and the interpreter of b-ICR is applied to the meta-interpreter of DCG top-down algorithm, b-ICR generates the knowledge about unknown words as hypothesis set.
1988-09-20