Research On Meta Search Engine Based On User’s Search Intention

Every day, network data is derived at the speed of what people can’t imagine, mass production and high speed transmission of network data, has made the existing information retrieval technology is not so good. Although the level of information retrieval has been greatly improved. However, facing the network data of explosive growth, the coverage of each retrieval tools, precision gradually decline. Under the background of huge amounts of data, the existing search engine has been completely unable to meet the users’ retrieval requirements of information retrieval system that "accuracy" and "efficiency".In order to get accurate information users have to spend a lot of time, repeatedly called different retrieval tools.To a certain extent, ease the contradiction of the "low-recall" of the single search engine when the Meta-search engine emerges. But in the form of a lengthy list of results for the user make they have to caught in the "query result overload" dilemma again.In order to effectively improve the "precision" and "recall" of retrieval system, the data mining, machine learning, artificial intelligence technology is applied in information retrieval, realizing the humanized and intelligent for retrieval tools.There is an important technical indicator of constructing a Chinese search engine: to offer semantic comprehension of query statements. That is, actual requirements or intention of users are given by semantic analysis of those query statements. The quality of query results could be greatly improved by offering information fetching service based on user query intention deduction.In construction of a Meta-search engine, it is of vital importance of make the procedures and techniques clear, in terms of information fetching and internal implementation. Also the trend of the field should be carefully reviewed. The main work of this thesis consists of: first, analysis of user intention, clearance of user query aim, extraction of search keywords, analysis of user query theme based on open directory tree, and offering of search engine schedule policy based theme relevance to users; second, compute global relevance based on voting theory; last, giving hierarchy cluster results based on improved suffix tree algorithm, while extracting class tags with information from occurrence frequency of feature words and user intention keywords, resulting in a recursive approach of hierarchy clustering.