自然言語理解システム　IMAGES-I　における構文処理について

概要

論文の詳細を見る
There are a lot of computer systems which interpret or understand natural language. Most of them are besed on the verb-centered case theory, and considerably successful in some restricted fields. In general, such restriction of fields brings decrease in syntactic and semantic complexities which are roughly proportional to the number of verbs used in the selected field. This has often resulted in ad hoc mixture of syntactic and semantic processings in the conventional systems. However, in order to construct some general and expalldable system, such mixture should be avoided. The authors have recently constructed an experimental understanding system named IMAGES-I (I__-nterlingual understanding M__-odel A__-iming at <GE>___-neral purpose S__-ystem-I). This system interprets sentences in an interlanguage which can realize still deeper semantic structures than the case theory, and moreover its syntactic and semantic processors are independent of each other. The syntactic processor transforms surface structures into surface dependency structures and transfers them to the semantic processor where semantic structures are generated from them and semantic anomalies, ambiguities and paraphrase relations are detected. This paper discusses the general performances of IMAGES-I and gives especially the details of its syntactic processing.
九州大学大学院総合理工学研究科の論文
1981-01-20