Discovering Homographs Using N-Partite Graph Clustering
This paper presents a method for discovering homographs from textual corpora. The proposed method first extracts an N-partite graph expression of word dependencies, and then, generates near-synonymous word clusters by enumerating and combining maximum complete sub-components on the graph. The homographs are identified as the words that belong to multiple clusters. In our experiment, we applied the method to Japanese newspaper articles and detected 531 homograph candidates, of which 31 were confirmed to be actual homographs.
KeywordsBipartite Graph Complete Bipartite Graph Compound Word Word Sense Disambiguation Multiple Cluster
Unable to display preview. Download preview PDF.
- 1.Hattori, S.: Iwanami course philosophy XI language. Iwanami Shoten, Tokyo (1968) (in Japanese)Google Scholar
- 3.Uno, T.: A Practical Fast Algorithm for Finding Clusters of Huge Networks. IPSJ SIGNotes Algorithms No. 088-001. Information Processing Society of Japan, Tokyo (2002) (in Japanese)Google Scholar