The Research On Knowledge-based And Graph-based Word Sense Disambiguation Algorithms

The development of network has brought explosive production of text information.Social networks,shopping platforms and so on are generating a large amount of text information every moment.Word ambiguity,as a common phenomenon in texts,brings the challenge in natural language processing.Although human beings can understand the meaning of ambiguous words well,the computers can not automatically recognize the meaning of words.As a result,in the fields of machine translation,information extraction,text categorization and other natural language processing,it is very difficult for all kinds of algorithms to understand the meaning of words.In order to solve the problems caused by word ambiguity,word sense disambiguation is proposed to improve the computer's understanding ability of ambiguous words.This dissertation focuses on the disambiguation algorithm based on knowledge and graph,and applies it in the short text classification.The main work is as follows:Firstly,we summarize the development process,classification and common external knowledge of word sense disambiguation algorithms,and compare the differences of different types of word sense disambiguation algorithms.Secondly,a semantic graph model based on global domain and short-term memory factor is proposed,which can better record global domain information and make disambiguation results more uniform.Meanwhile,word sense disambiguation based on the semantic graph improves the ability of word sense disambiguation to use global and local semantics.A large number of experiments show that the improved semantic graph model can significantly improve the results of word sense disambiguation algorithm.Finally,in order to solve the instantaneity,non-standard and semantic sparsity of short text,word sense disambiguation based on graph and external knowledge is applied to short text classification,so that the new method can effectively enhance the semantic density of short texts.Experiments show that the proposed method improves the classical algorithm to a certain extent,which plays a good effect in short text classification using word sense disambiguation.