Generation and evaluation of word embedding in each sense using word sense disambiguation

Abstract

In recent years, word2vec is used to generate word embedding which is one of the important techniques in natural language processing. It can generate word embedding that represents the meaning of words based on surrounding words in corpus. Since there are polysemic words that can be used to express more than one meaning, it is desirable that we can generate word embedding in each sense. In this paper, we propose a method of generating word embedding in each sense by word sense disambigu- ation of words in large-scale corpus realized by machine learning using corpus with semantic tags. An evaluation experiment through a questionnaire showed that suitable word embedding in each sense was generated.