Discovering and Using Groups to Improve Personalized Search

Proceedings of Web Search and Data Mining (WSDM) 2009 |

Published by Association for Computing Machinery, Inc.

Personalized Web search takes advantage of information about an individual to identify the most relevant results for that person. A challenge for personalization lies in collecting user profiles that are rich enough to do this successfully. One way an individual’s profile can be augmented is by using data from other people. To better understand whether groups of people can be used to benefit personalized search, we explore the similarity of query selection, desktop information, and explicit relevance judgments across people grouped in different ways. The groupings we explore fall along two dimensions: the longevity of the group members’ relationship, and how explicitly the group is formed. We find that some groupings provide valuable insight into what members consider relevant to queries related to the group focus, but that it can be difficult to identify valuable groups implicitly. Building on these findings, we explore an algorithm to “groupize” (versus “personalize”) Web search results that leads to a significant improvement in result ranking on group-relevant queries.