A Colorado State University professor and graduate student have teamed up to create QueryTracker, a software program that improves long-term Web searches by reducing irrelevant search returns while also helping to overcome poorly formed queries.
Computer Science Professor Adele Howe, and Gabriel Somlo, a Colorado State graduate student, developed the continuous query software program to be installed on top of a conventional search engine, such as Google, to actively exploit information about a user’s recurring interests. QueryTracker automatically generates and submits additional queries each day based on what it learns about a user’s interests and priorities over time. QueryTracker then filters the results, finds the information relevant for the specific individual, and sends that information to the user.
"The program exploits feedback from the user to refine its idea of what the user is really interested in instead of relying on the typical two to three word query to capture a complex information need," said Howe. "QueryTracker augments the original query based on this knowledge and sends that to the search engine. Additionally, the program filters what is returned to remove pages that have been seen before or that don’t appear to be relevant."
When reviewing results, QueryTracker asks users to enter feedback on which links are or are not relevant to their queries. QueryTracker uses this feedback to build a query profile which helps improve relevance predictions for new documents encountered in the future.
The proof of concept program is currently in testing stages. However, Web users can go to the QueryTracker home page at www.cs.colostate.edu/~somlo/QueryTracker/ to learn more about the program and register to try the program as part of the testing process.