Argument Status in Japanese Verb Sense Disambiguation

This research aims to incorporate argument status-based modelling within an otherwise selectional constraint-based system of verb sense disambiguation, to capture effects such as underspecification, surface case alternation and semantic backingoff. The proposed implementation hinges around a description of the general behavioural characteristics of integral complements, complements, middles and adjuncts through a pre-determined weighting schema. On limited evaluation, the resultant system returned an accuracy of over 83%, and was further shown to significantly outperform baseline methods.