A Word Sense Probabilistic Topic Model

This paper proposed a novel probabilistic topic model based on word senses. Different from the classic topic model exploring word form, this model generated the word form and at the same generated the word sense in a specific context. There are totally four layers in this model compared with the three layers in transitional probability topic models. We further illustrated how to solve the parameters for this word sense probabilistic topic model (WSPTM). As far as the applications of WSPTM are concerned, a basic task for natural language processing, i.e. word sense disambiguation would be benefited. We further illustrated how to solve the parameters for this word sense probabilistic topic model.