Definitive results finding in web search environment
Yong-Mei Gao, Yalou Huang · 2009
In web search area, queries with definitive results are frequently issued by people, such as “163” whose definitive URL is “http://www.163.com”. The definitive relationship of these queries and corresponding pages are much more stable than that of other queries. Improving search result for queries of this kind is expected to make a beneficial effect on overall performance of a search engine. According to this problem we proposed a search framework composed of online and offline two cooperative parts. Offline processor is an offline ranker used to automatically extract definitive ≪query, page≫pairs which will be seen as descriptions for corresponding pages and then given higher weights in online process to insure a better search result. ≪anchor text, destination page≫ pairs are used as candidate definitive pairs, because anchor texts can be seen as latent queries and can reply future issued queries. Above all, this framework moves the key point to offline part so it can be solved without critical limits on time and space and can meet online users' requirements. Experiment results showed that definitive results got by offline ranker can obtain 0.8647 on precision@top1 for 1400 queries on 100M URL corpus. A big promotion was made compared with traditional BM25_content based method (0.5845).