The Valence Patterns of Japanese Verbs Extracted From The EDR Corpus

This paper describes research on particular verb valences obtained from actual linguistic data. We created verb valence data using data from the EDR Co-occurrence Dictionary as our source. The EDR Co-occurrence Dictionary is coded with syntactic governing-dependent relation tags and semantic tags. The syntactic governing-dependent relations data in the EDR Co-occurrence Dictionary however, is expressed as individual constituent pairs. In this study, we grouped each of the governingdependent relation pairs according to their verb concepts and then unified them into a number of combinations based on case. After the data was automatically unified from the source data, we manually corrected mistaken governing-dependent relations, and also made changes to case where necessary. By following this procedure, we created basic valence data for each verb. Further, based on this valence data and the verb patterns created from it, we are currently looking into creating semantic groups for nouns on which semantic restrictions are imposed by the verb.