대화체 기계번역을 위한 사전의 통사 , 의미 정보

Abstract

This paper aims to examine kinds of syntactic and semantic information to be encoded in lexical entries of the lexicon which should ulticately be used for machine translation of spoken languages. The reason why we need as detailed information as possible is that in the real situation of speech, unlike the case of texts, it is difficult to get reliable phonological information. Among needed syntactic information, detailed subcategorization information is necessary. For Korean nouns, it is unavoidable to have the distinction of mass and count nouns, as in English. One kind of useful semantic information can be encoded with semantic features, by means of which selectional restriction may be handled well. Another way of representing syntactic and semantic information is using Mel`cˇuk`s lexical relations used in his Explanatory Combinatorial Dictionary. Lexical relations give a promising method to handle collocation, which should be of prime concern in any study of (machine) translation.