Extraction of Synonymous Relationships Using Japanese-English Bilingual Titles of Agricultural Academic Papers

Abstract

We have developed agricultural linguistic resources (such as the Japan Agricultural Thesaurus, JAT) to process electronic data effectively. Here, we report a method for extracting synonymous relationships among agricultural terms by using Japanese-English bilingual titles of academic papers. We first collect bilingual term pairs with a high co-occurrence frequency from parallel titles, and manually check to verify that they are parallel translations. Then we generate a network graph representing terms as nodes, and lines connecting translation pairs as links. The network graph is composed of many kinds of disjointed sub-graphs. The number of sub-graphs with more than two terms is 961. All terms included in a same sub-group are determined to have a synonymous relationship.