Research On Query Expansion Based On Latent Semantic Analysis

Query expansion is a key technology of information retrieval research, and itis an important way to improve information retrieval efficiency. In informationretrieval, users often need to try many times to construct the query to the desiredtarget literature because of the diversity of user knowledge expression differencesand retrieval environment. Traditional query expansion method focuses onsynonymous substitutions in each of the original query words, but in practice it isdifficult to adapt to “polysemy” and “synonym”. Latent semantic analysis is amethod of knowledge acquisition and expression, collecting latent semanticstructure of lexical items by using the statistical analysis method, then thesemantic documents are closer in multidimensional space after matrix operation.The latent semantic analysis technique is applied to query expansion process, andit can improve retrieval efficiency for better express the original query.This paper introduces the domestic and international situation of queryexpansion at first, then raises the main problem that query expansion is unable toaccurately express semantic information query through comparing the advantagesand disadvantages of various methods. The article introduces the latent semanticintelligent retrieval, explains its background and basic principle, combinessemantic dictionary extended dominance and probabilistic latent semantic, raisesa new method of query expansion. At last, the paper proves the feasibility andeffectiveness of the method with a sample test.The main content is:(1)The paper analyzed the necessity of query expansion research,summarized the research status, and finally pointed out the limitations of existingquery expansion method from the comparison and analysis of the computationalcomplexity, retrieval efficiency at different extension methods advantages anddisadvantages. (2) The paper introduced the semantic query expansion method based onsemantic dictionary analyzed its effectiveness and convenience. This method isthe basic theory of query processing.(3)The paper introduced the theoretical basis of latent semantic analysis,singular value decomposition. The author thought that the latent semanticanalysis could reduce the effect of synonyms and ambiguity search extensions.The test confirmed the latent semantic analysis for query expansion with a smallsample set, also analyzed the deficiencies can be improved.(4)To explore latent semantic analysis based query, the article proposed amethod that connected probabilistic latent semantic analysis with semanticdictionary. This method got ideal precision and recall after query expansion in thecluster expansion of text clustering after comparing with the existing methods.The method will be applied to more information retrieval models accordingto this study in the future.