Discrimination of semantic depiction and statistical word sense

In order to improve the effect of word sense discrimination,to obtain substantial improvement,we need to look at the relationship between semantic characterization and word sense understanding,need to research these problems by the theore and experiment. This article analyzes the impact of various factors that characterize word sense in the view of Linguistics.Furthermore,we design and implement the comparative experiments between human and machine word sense discrimination.Those experiments reveal the relationship between granularity of semantic depiction and accuracy of word sense discrimination.Then it shows the necessity of adjusting the size of semantic characterization.Experimental results show that scientifically controls the granularity of semantic depiction can enhance the computability of word senses,and improve the accuracy of word sense discrimination.