Japanese Information Retrieval Method Using Syntactic and Statistical Information

This paper presents a Japanese information retrieval method using the dependency relationship between words and semantic and statistical information about them. Our method gives a score to each document to be retrieved, based on a probabilistic model, giving an additional score based on the dependency relationship between words, and ranks the documents according to their score. In order to make use of semantic information, our method performs query expansion with a thesaurus, and zero pronoun resolution. Experimental results show that our method is more effective than a method using linguistic and statistical information independently.

• Type: Conference
• Year: 2001
• Where: NLPRS
• Authors: Tsunenori Mine, Hiroki Fujitani, Makoto Amamiya
• Keywords: Dependency Relationship, Information Retrieval Method, Method Performs Query, Natural Language Processing, NLPRS 2001