Using query logs and click data to create improved document descriptions
Maarten van der Heijden, Max Hinne, Wessel Kraaij, Suzan Verberne, Theo van der Weide · 2009
Logfiles of search engines are a promising resource for data mining, since they provide raw data associated to users and web documents. In this paper we focus on the latter aspect and explore how the information in logfiles could be used to improve document descriptions. A pilot experiment demonstrated that document descriptors extracted from the queries that are associated with documents by clicks provide useful semantic information about documents in addition to document descriptors extracted from the full text of the web pages.