Socially filtered web search: an approach using social bookmarking tags to personalize web search

Today's knowledge workers are confronted with an ever increasing information overload while searching for needed information in the web. Common search engines do not take into account the current work context of the user. But we consider context information as an effective means to implicitly narrow the information space of the web. In this paper we present a novel approach that increases the relevance of search results by considering the current work context. We track the user's web browsing behavior, store visited pages and build up a user model based on this information. As the user browses, the stored URLs of the visited pages are enhanced with tags from social bookmarking sites. Based on the user model and the retrieved bookmarks we developed an easy-to-use and easy-to-configure clientside web search engine that refines the original search query with these tags. Our approach follows the design principle of non-intrusiveness. That means we present the context-sensitive personalized adapted search results together with the original non-adaptive search results. We developed an open architecture that allows the user to reconfigure the system to use different metadata providers and search engines. In order to prove our architecture we implemented a Firefox Add-on.