QueryTracker: An Agent for Tracking Persistent Information Needs

Most people have long term information interests. Current Web search engines satisfy immediate information needs. Specific sites support tracking of long term interests. We present an agent that satisfies a gap in these services. QueryTracker implements a search engine interface with state. A user’s query and a learned alternative is automatically submitted daily to a search engine. A profile of the user’s interest is constructed based on user relevance feedback. Daily search results are disseminated if they correspond closely to the user’s profile. Changed pages are checked for relevant additions. We evaluate the impact of generating the alternate query and assess alternative methods of filtering the results in a month long user study. Results show that QueryTracker is effective at disseminating relevant documents over a period of time.

• Type: Conference
• Year: 2004
• Where: ATAL
• Authors: Gabriel Somlo, Adele E. Howe
• Keywords: ATAL 2004, Long Term, Long Term Interests, Search Engine