Incremental methods for context-basedWeb retrieval

Abstract

Intelligent search depends on effective methods for identifying the information needs of a user and making relevant information resources available when needed. Reflecting user context has long been recognized as a key aspect to realizing the potential of intelligent Web search. This paper proposes a theoretical basis for better understanding the role of context in Web retrieval. It addresses the problem of identifying context-specific terms, finding relevant information sources, and automatically formulating and refining queries. We describe ongoing research on the use of incremental methods to retrieve relevant content through two main approaches. The first, feed-based, periodically checks for new relevant items in specific websites by accessing RSS feeds. The second, query-based, incrementally formulates queries, which are submitted to search interfaces (e.g., major search engines or individual search forms). We discuss the technical challenges imposed by these approaches, outline our system architecture, and present preliminary evaluations of the proposed techniques.