Discovering Single-Query Tasks from Search Engine Logs

Abstract

When a user uses a search engine to find out some information, for each query that he does, search engine gives some links as result. Search engines provide these results for the user and on the other hand save the log information of each user about queries have done by them. These log information through query mining methods and algorithms are used to extract useful knowledge. In previous works a concept has defined as “task” which is a set of possibly noncontiguous queries which refer to the same need of information. Clustering methods have been used for discovering collective tasks by aggregative similar user tasks, possibly performed by distinct users. All these studies were taking into account just the queries which have done in a session and then concept of mission came up and it was trying to find multi tasks even in different sessions. Well, till now all studies were trying to cluster and aggregate the queries which are done for same information need and none of them cares about queries which are done independently and without any relation to other queries. Queries that we can call them singleton or single task queries. We thought that may be finding these queries could have some more interesting benefits for search engine either by taking them off and eliminating them or studying more on them and improving the results of search engines about responding to these kind of queries. Our contribution is using classification methods to find and classify these single task queries than multi tasks. Based on saved information about query logs we tried to define features for a single query and use these features in classification algorithms to achieve to our goal.