Topic Detection And Tracking Based On Semantic Framework

Query Expansion is an important branch of information retrival, which has been an effective measure to help us using search engines efficiently. With the complexity and diversification of Internet information, especially the raid development of microblog and WeChat, the traditional algorithms of query expansion are not able to suggest right keywords in nonstandard short text. Because of term independence assumption and information loss in traditional retrieval model, we are not able to get sufficient semantic information in the process of query expansion, the it is unable to solve synonyms and polysemant in natural language processing.This paper mainly focuses on information retrieval model and query expansion. Based on comparison of some different models, we find that the main reason for above problem is the lack of effective semantic analysis. We propose a method of query expansion based word activation forces to improve the quality of expansion terms.The main contribution of this paper is stated as below:Firstly, this paper implemented experiment to find out the difference between traditional term similarity algorithm and WAR The result proved that the word networks which have been built by WAF can capture the key words in a topic.Secondly, according to the characters of short text, this paper proposed a new computational method of word activation force which introduced the relationship between sentences to reduce interferences caused by nosiy feature. To improve the quality of expansion terms, we use the distribution of word affinity to adjust the sequence of associated words.Thirdly, this paper proposed a new query expansion algorithm based on the combination of word network model and topic clustering, which has been integrated to microblog processing system. The contrast experiment between traditional query expansion based on BM25and our algorithm shows the great advantage made by the semantic model of WAR...