Utility-based repair of inconsistent requirements

Abstract

Knowledge-based recommender applications support the customer-individual identification of products from large and complex assortments. Recommendations are derived from customer requirements by interpreting filter constraints which reduce the set of possible products to those relevant for the customer. If no solution could be found for the requirements, repair actions are proposed which support customers in finding a way out of the "no solution could be found" dilemma. State-of-the-art systems support the identification of repair actions based on minimality assumptions, i.e., repair alternatives with low-cardinality changes are favored compared to alternatives including a higher number of changes. Consequently, repairs are calculated using breadth-first conflict resolution which not necessarily results in the most relevant changes. In this paper we present the concept of utility-based repairs which integrates utility-based recommendation with efficient conflict detection algorithms and the ideas of model-based diagnosis (MBD). © 2009 Springer Berlin Heidelberg.