A Method for Extraction of Similar Words from Japanese Compound Nouns based on Bipartite Graph

Abstract

　This paper proposes a method for similar words groups extraction from nouns in a Japanese corpus. This method is based on maximal complete bipartite graph included bipartite graph made from compound nouns. In this bipartite graph, one node set tends to be a similar words group and the other node set to be a group of view point words, when the graph is maximal complete. Further, this method can give also similarity measurements between words using relations among those similar words groups.