Bi-Matching Mechanism to Combat the Long Tail of Word Sense Disambiguation

Anonymous

Abstract: The long tail phenomenon of word sense distribution in linguistics causes the Word Sense Disambiguation (WSD) task to face a serious polarization of word sense distribution, that is, Most Frequent Senses (MFSs) with huge sample sizes and Long Tail Senses (LTSs) with small sample sizes. The single matching mechanism model that does not distinguish between the two senses will cause LTSs to be ignored because LTSs are in a weak position. The few-shot learning method that mainly focuses on LTSs is not conducive to grasping the advantage of easy identification of MFSs. This paper proposes a bi-matching mechanism to serve the WSD model to deal with two kinds of senses in a targeted manner, namely definition matching and collocation feature matching. The experiment is carried out under the evaluation framework of English all-words WSD and is better than the baseline models. Moreover, state-of-the-art performance is achieved through data enhancement.