Actively Mining Search Logs for Diverse Tags

Social tagging has become a very important mechanism for organizing information on the Web. Usually, people tag a web page manually, just as what they do on a social bookmarking web site. In this paper, we will demonstrate a brand-new perspective - tagging web pages automatically by mining search logs. In order to keep diversity, we first classify web queries into different categories and then extract tags from queries to depict each category. Thereafter we describe a web page with all queries which are related to this page, and finally we get the recommended tags for each web page after mapping the related queries into corresponding diverse tags. The experiments conducted on a real search log show that our method can dig out accurate and meaningful diverse tags for web pages more effectively.