Analysis of Korean Compound Noun using Lexical Information between Nouns

Compound noun analysis is a difficult problem because the relationship between noun components depends on lexical meaning. This paper presents a method to analyze the structures of nominal compounds based on the linguistic relations between nouns and their lexical co-occurrence relations which are extracted from the corpus. Compound noun includes a sequence of nouns and noun phrase modified by a noun with adnominal postposition. Two nouns in a compound noun are linked by either the predicate-argument or the qualifier-head relation at the syntactic level. The two relations are obtained from the corpus and applied to nominal compound analysis. Lexical co-occurrence data were extracted by the POS tagger and the partial parser from 30 million words of Yonsei Lexicographical Center Corpus. The precision rate of analysis is 83.8% for compound nouns selected from the test corpus separated from the training cropus.