Research On Intelligent Query Expansion Technology Based On Users’ Feedback

In recent years, with the development of Internet technology, it has become one of the most important methods for people to acquire information. However, the existence of massive, heterogeneous and complex information makes it difficult for people to acquire information. As an important branch of Internet information technology, search engine can make people acquire information more effectively. Although the existing search engines are able to satisfy users’ retrieval requirements, there still exists some problems to be optimized. For example, users may not organize the query well to express retrieval demand accurately, which results in the mismatch between query and document index, and results in low accuracy of retrieval results. The query expansion technique studied in this thesis is an effective way to solve this problem. The main work includes the following four aspects:Analyzed some relevant query expansion algorithms, such as query expansion algorithms based on LSA, Ontology, ESA, Rocchio, Ide, Ide-dec, LCA, LA, LOCOOC, Apriori, NNLM, KLD, CHI1, CHI2 technology, as well as the fusion analysis of different algorithms, the design of weight combination, the optimization of each parameter etc.Through the research on the relevant query expansion algorithms, this thesis proposed a hybrid query expansion algorithm based on the reranking of documents. Firstly, on the basis of feedback information for initial retrieval, this method adopted the proposed strategy to rerank the retrieved documents, and then used the hybrid query expansion algorithm to analyze the top n reranked documents.By carrying out multigroup contrast experiments on OHSUMED dataset, we validated the effectiveness of the proposed documents reranking based hybrid query expansion algorithm, and it can improve the performance of retrieval system effectively. Compared with the existing query expansion algorithms, th is method can achieve a better performance in terms of the mean average precision MAP.Applied the method to some existing search engines. It was mainly based on the Google and Bing’s open API. Firstly, users should provide the feedback information interactively, and then adopted the method proposed in this thesis to do query expansion analysis, finally expanded some words to the origin query and executed the second retrieval automatically.