ReAction: Personalized Minimal Repair Adaptations for Customer Requests

摘要

Knowledge-based recommender systems support users in finding interesting products from large and potentially complex product assortments. In such systems users continuously refine their specifications which the product has to satisfy. If the specifications are too narrow no product can be retrieved from the product assortment. Instead of just notifying the customer that no product could be found we introduce an approach called ReAction to support customers with minimal repair adaptations. In this paper we give a detailed explanation of our algorithm. Besides that we present the results of a detailed empirical evaluation focussing on the quality as well as on the runtime performance. The work presented is relevant for designers and developers of database systems as well as knowledge-based recommender systems interested in (ⅰ) identifying relaxations for database queries, (ⅱ) applying and dealing with user utilities, and (ⅲ) improving the system usability through suggesting minimal repair adaptations for inconsistent queries.