Conflict-Directed Relaxation of Constraints in Content-Based Recommender Systems
Content-based recommenders are systems that exploit detailed knowledge about the items in the catalog for generating adequate product proposals. In that context, query relaxation is one of the basic approaches for dealing with situations, where none of the products in the catalogue exactly matches the customer requirements. The major challenges when applying query relaxation are that the relaxation should be minimal (or optimal for the customer), that there exists a potentially vast search space, and that we have to deal with hard time constraints in interactive recommender applications.
In this paper, we show how the task of finding adequate or customer optimal relaxations for a given recommendation problem can be efficiently achieved by applying techniques from the field of model-based diagnosis, i.e., with the help of extended algorithms for computing conflicts and hitting sets. In addition, we propose a best-effort search algorithm based on branch-and-bound for dealing with hard problems and also describe how an optimal relaxation can be immediately obtained when partial queries can be (pre-)evaluated.
Finally, we discuss the results of an evaluation of the described techniques, which we made by extending an existing knowledge-based recommender system and which we based on different real-world problem settings.
Unable to display preview. Download preview PDF.
- 1.Bridge, D.: Product recommendation systems: A new direction. In: Weber, R., Wangenheim, C. (eds.) Proceedings of the Workshop Programme at the Fourth International Conference on Case-Based Reasoning, pp. 79–86 (2001)Google Scholar
- 2.Console, L., Friedrich, G., Dupré, D.T.: Model-Based Diagnosis Meets Error Diagnosis in Logic Programs. In: IJCAI 1993, Chambéry, France, pp. 1494–1501 (1993)Google Scholar
- 3.Felfernig, A., Friedrich, A.G., Jannach, D., Stumptner, M.: Hierarchical Diagnosis of large configurator knowledge bases. In: McIlraith, S., Dupré, D.T. (eds.) 12th Intl. Workshop on Principles of Diagnosis, pp. 55–62 (2001)Google Scholar
- 5.Freuder, E., Wallace, R.J.: Partial Constraint Satisfaction. Artificial Intelligence (58) (1992)Google Scholar
- 7.Jannach, D.: ADVISOR SUITE - A knowledge-based sales advisory system. In: Proceedings of ECAI/PAIS 2004, Valencia, Spain, pp. 720–724. IOS Press, Amsterdam (2004)Google Scholar
- 8.Junker, U.: QUICKXPLAIN: Preferred Explanations and Relaxations for Over-Constrained Problems. In: Proceedings AAAI 2004, San Jose, pp. 167–172 (2004)Google Scholar
- 9.McSherry, D.: Explanation of Retrieval Mismatches in Recommender System Dialogues. In: ICCBR 2003 Mixed-Initiative Case-Based Reasoning Workshop, Trondheim, Norway (2003)Google Scholar
- 12.Mittal, S., Falkenhainer, B.: Dynamic constraint satisfaction problems. In: Proceedings of the 8th National Conference on Artificial Intelligence, AAAI 1989, pp. 25–32 (1989)Google Scholar
- 14.Schiex, T., Fargier, H., Verfaille, G.: Valued constraint satisfaction problems: Hard and easy problems. In: International Joint Conference on Artificial Intelligence IJCAI 1995, Montreal, Canada, pp. 631–639 (1995)Google Scholar