Research And Application Of Topic Oriented Text Mining

With the continuous development of the informatization of scientific research management,many research documents have been accumulated.These documents contain abundant information but have not yet been fully utilized.On the one hand,researchers cannot retrieve documents efficiently and precisely.On the other hand,researchers need a tool to analyze the document sets in a whole view from a certain aspect.To solve the problem,this thesis researched topic-oriented text mining technology,extracted out the topical information in research documents,and implemented a topic-based search and analysis system for research documents.The system enables researchers to make use of existing research resources conveniently and efficiently.This thesis conducted a thorough survey on topic-oriented text mining technology,proposed a method for constructing word similarity matrix based on word embedding and Tongyici Cilin(a Chinese thesaurus),and evaluated the modeling result of Graph sampling algorithm with this matrix.The experimental result shows that the word similarity matrix based on word embedding and Tongyici Cilin well expresses the lexical semantic similarities and using Graph sampling algorithm with this matrix while sampling topics can effectively improve the results of topic-oriented text mining.Moreover,this thesis conducted an in-depth analysis of researchers' needs for searching and analyzing research documents according to topics.Based on these needs,a reasonable module design and architecture design are carried out,and a topic-based search and analysis system for research documents is implemented,which provides functionalities including document search,document recommendation and topic trend analysis.This thesis researched and implemented research documents retrieval and analysis according to topics,enriches the methods of searching research documents and increases the efficiency of analyzing research documents,has good research and application value.