Towards the automatic classi¯cation of Japanese verbs

This paper is the ¯rst of a series of papers dealing with the automatic classi¯cation of verbs and the automatic determination of the values of a verb's semantic features. Both tasks are in fact identical, as verb classes are nothing more than clusters of verbs with similar constellations of semantic features. Thus, being able to automatically determine a verb's class will enable us to pre-dict the constellation of semantic features inherent to that verb. Likewise, knowing the values of a verb's semantic features enables us to accurately predict a verb's class. Ultimately, the aim of this series is to arrive at a e±cient algorithm for classifying Japanese verbs. In order to achieve this goal, each paper in the series will pick up an article published by the leading journals in the ¯eld, and evaluate it's bene¯ts and drawbacks compared to other papers on the subject. The ¯rst article discussed is [Merlo and Stevenson 2001]