Research And Application Of Extracting Semantic Relations On Minning Query Logs

In current century,information bomb becomes remarkable with a high-speed update,and users' requirements about search results continues increasing,so that how to achieve useful information from a huge mount of web information resources is one of the vital problems. Query recommendation is useful to modify queries for search again.On one hand, it can solve the problem of query can not exactly describe the user's intent. On other hand, it can solve the problem of the ambiguity of language reduce the precision rate of the search result.An extracting semantic relations algorithm (ESR) from Query Logs is presented.First construct a weighted Query-URL bipartite graph from query log data. Then compute the semantic similarity of queries by distance of queries nodes and synonymy similarity and use a effective mining algorithm to discovering semantic related queries based on graph path.Experiments show that the average precision rate of the algorithm can get 0.825 and increase by 2.5% compared with substring extending algorithm.A patent search system based on query recommendation and term is presented by analyzing the character of the patent data. The system is evaluated by Chinese Web Test collection with 200 GB web pages (CWT200g).Experiments show that the average R-P,the average P@10,the average Bpref of the system respectively get 0.105,0.173,0.081,and respectively increase by 23.02%,20.44%,18.88% compared with the system without term. Experiments based on the one month query logs which are generated by the system running on Hunan Intellectual Property Office,shows that the average score of query recommendation get 3.91 by 5-point scale.