拡張ユニフィケーションを用いたパーザIPの実現手法

概要

論文の詳細を見る
This paper presents an extended Prolog and its application to an integrated parser for text understanding. The word "integrated" includes some meanings. First, syntactic, semantic, and contextual analyses occur as an integral part of the parsing process. Second, three distinct metaphors available in the field of computational linguistics, such as procedure oriented, declaration oriented, and actor oriented metaphors are incorporated into a single grammar formalism. Third, two major discourse analyses, prediction-driven and explanation-driven approaches are unified into a single module. In this paper, we concentrate on the implementation technique of the integrated parser. The syntax of the integrated parser is based on a Lexical Functional Grammar proposed by Bresnan, which was grown out of ideas from current computational linguistics and transformational linguistics. The language Prolog discussed here has been extended by allowing the inclusion of both "equality assertions" proposed by Kornfeld and computational entities "actors" proposed by Hewitt. The notion of "equality assertions" is well suited for defining new data types, for implementing the parser which is independent of the physical representation of its internal data structure, and for delaying the evaluation of grammatical constraints until the necessary information is acquired. The notion of "actors" is introduced to perform parsing via "message passing". The actor-oriented parsing offers simplicity. modularity, and extensibility. It is quite natural and general to implement the Lexical Functional Grammar in the extended Prolog. Moreover, the added features, a new data type and actor-oriented paradigm, do not impinge on the efficiency of the integrated parser.
社団法人人工知能学会の論文
1986-09-01

拡張ユニフィケーションを用いたパーザIPの実現手法

スポンサーリンク

概要

著者

関連論文

スポンサーリンク