To users with unique individual tastes libra is an

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: is an initial contentbased book recommender which uses a simple Bayesian learning algorithm and information about books extracted from the web to recommend titles based on training examples supplied by an individual user. Initial experiments indicate that this approach can e ciently provide accurate recommendations in the absence of any information about other users. Libra includes in its content descriptions information on related books and authors obtained from collaborative methods, and this information was experimentally shown to positively contribute to its performance. In many ways, collaborative and content-based approaches provide complementary capabilities. Collaborative methods are best at recommending reasonably well-known items to users in a communities of similar tastes when su cient user data is available but e ective content information is not. Contentbased methods are best at recommending unpopular items to users with unique tastes when su cient other user data is unavailable but e ective content information is easy to obtain. Consequently, as discussed above, methods for further integrating these approaches can perhaps provide the best of both worlds. Finally, we have discussed problems with previous recommender evaluations that use commercial data in which users have selected their own examples. Although there are good and bad aspects of all existing evaluation methods, we have argued for the advantages of using randomly-selected examples. 6 ACKNOWLEDGEMENTS Thanks to Paul Bennett for contributing ideas, software, and data, and to Tina Bennett for contributing data. This research was partially supported by the National Science Foundation through grant IRI-9704943. References 1 T. Anderson and J. D. Finn. The New Statistical Analysis of Data. Springer Verlag, New York, 1996. 2 S. Baker. Laying a rm foundation: Administrative support for readers' advisory services. Collection Building, 123-4:13 18, 1993. 3 M. Balabanovic and Y. Shoham. Fab: Content-based, collaborative recommendation. Communications of the Association for Computing Machinery, 403:66 72, 1997. 4 C. Basu, H. Hirsh, and W. W. Cohen. Recommendation as classi cation: Using social and content-based information in recommendation. In Proceedings of the Fifteenth National Conference on Arti cial Intelligence, pages 714 720, Madison, WI, July 1998. 5 D. Billsus and M. Pazzani. A personal news agent that talks, learns and explains. In Proceedings of the Third International Conference on Autonomous Agents, Seattle, WA, 1999. 6 D. Billsus and M. J. Pazzani. Learning collaborative information lters. In Proceedings of the Fifteenth International Conference on Machine Learning, pages 46 54, Madison, WI, 1998. Morgan Kaufman. 7 M. E. Cali and R. J. Mooney. Relational learning of pattern-match rules for information extraction. In Proceedings of the Sixteenth National Conference on Articial Intelligence, Orlando, FL, July 1999. 8 C. Cardie. Empirical methods in information extra...
View Full Document

This document was uploaded on 09/12/2013.

Ask a homework question - tutors are online