자연언어처리를 위한 관용표현 연구 = A Study on Idiomatic Expression for Natural Language Processing

다국어 초록 (Multilingual Abstract)

Kim, Han-Saem. 2003. A Study on Idiomatic Expression for Natural Language Processing. Korean Semantics, 13. In natural language processing, one of the most significant problem is ambiguity in lexical meaning. Sometimes ambiguity in meaning and structure of sentence is caused by idiomatic expression. The efficient processing of it requires inductive researches based on large corpora. Due to lack of complete description of idiomatic expression the performance of latest system is not so satisfactory yet. Idiomatic expression is called MTWU (Multi-Word Translation Unit) in field of machine translation because it is treated as a single unit in MT. In a broad sense idiomatic expression, concord relation, and sentence pattern belong to MTWU. For solving the problem in idiomatic expression it is needed to construct the database deep into several kinds of information. Because there is no relation between th frequency of idiomatic meaning and literal meaning the frequency information is very important. Idiomatic expression is distinguished from general phrase with information about modifier and contiguity of each component. Some idiomatic expression can be discriminated by its frozen form. Also fundamental syntactic and semantic features of each idiomatic expression are important information in processing. Through the objective and accurate database technology in NLP can have great possibility.