User Lenses — Achieving 100% Precision on Frequently Asked Questions
The concept of a “user lens” is introduced. The lens is a sequence of linear transformations used to reweight the vectors which represent documents or queries in information retrieval systems. It is trained automatically via relevance data provided by the user. Experiments verify the lens can improve performance on training data while not degrading test data performance, and that larger lenses result in nearly perfect performance on the training set. The lens provides a mechanism for automatically capturing long-term, user-specific information about an improved representation scheme for document vectors.
KeywordsWeighting Scheme Relevance Feedback Ranking Score Information Retrieval System System Verification
Unable to display preview. Download preview PDF.
- Bartell, B. T. (1994). Optimizing Ranking Functions: A Connectionist Approach to Adaptive Information Retrieval, thesis, Department of Computer Science and Engineering, The University of California, San Diego, CSE 01 14, La Jolla, CA 92093.Google Scholar
- Harman, D. K., ed. (1997). The Fifth Text REtrieval Conference (TREC5). Gaithersberg, MD: National Institute of Standards and Technology. NIST Special Publication 500–238.Google Scholar
- Salton, G., ed. (1971). The SMART Retrieval System — Experiments in Automatic Document Retrieval. Englewood Cliffs, N.J.: Prentice-Hall Inc.Google Scholar
- Vogt, C., Cottrell, G., Belew, R., and Barteil, B. (1997). Using relevance to train a linear mixture of experts. In Harman, D. K., ed., The Fifth Text REtrieval Conference (TREC5), 503–515. Gaithersberg, MD: National Institute of Standards and Technology. NIST Special Publication 500–238.Google Scholar