Mining Search Engine Query Logs for Query Recommendation

• Zhiyong Zhang, SINA Corporation, China
• Olfa Nasraoui, University of Louisville, USA

Full text:

Track: Posters

This paper presents a method of mining search engine query logs to get query recommendations. In order to get a more well-rounded solution, the authors try to combine two methods together. On the one hand, the authors study and model search engine users' sequential searching behaviors closely. They interpret this consecutive search behavior as a client-side query refinement, which is in comparison with the search engines' own query refinement process, and take fully advantage of this process to acquire some useful information that helps to generate the related queries. On the other side, the authors combine this method with the traditional content based similarity method. Also, some tentative method were devised to clean the query log and prune the robots' visit log. To experiment and evaluate our method, we use about one hundred days practical query log from SINA' search engine1 to do off-line mining. And we invite three independent editors from SINA to evaluate our result. Based on their subjective judgement, our method was very effective for finding the related queries.

Other items being presented by these speakers

• Focused Crawling: Experiences in a Real World Project (Posters Track)