The Study Of Concept Mining In Information Retrieval System Based On Concept Lattice

"The system of automatic query expansion based on concept lattice" (AQECL) has the different way from the traditional method of query expansion. AQECL attempts to use the technology of text concept mining, text concept relation, and the algorithm of concept lattice construction to provide automatic query expansion from the concept point of view.Followed the theory of formal concept analysis (FCA), this thesis will focus on the algorithm of text concept extraction, which is one of the most important steps in AQECL. With the basis of concept ,the center of original query, and the active modification, AQECL can provides all-around and clear suggestions to users. Major works include:1, A new module of query expansion is added to the traditional IR system. Following FCA and the application direction of concept lattice, a module of query expansion, based on concept lattice is designed and realized. This new module will provides the way of query expansion via the construction of text concept relation. At the same time, the new module can also provide the Hasse graphics, which will improve the exchange between users and our IR system.2, The focus of this thesis is text concept extraction, and a demo system for the preprocess module is implemented in AQECL, and original testing has been finished. The concept of term entropy (TE), from information entropy point of view, is used to evaluate term weigh, instead of the traditional IDF. Original testing has proved that, the method of TE can be compared to CHI; however, TE will improve the computing efficiency to some extent.3, At the same time, knowledge background of domain lexicon is the added after the preprocess module to make the term weight is correlative to time. Additionally, the structure information of Web text is also been taken into account. From the results of original testing, these attempts can improve the precision of concept extraction without expense of efficiency.