Recommender Systems: Knowledge from Mining User Experiences

Résumé du livre

Abstract: "One largely untapped source of knowledge about large data collections is contained in the cumulative experiences of individuals finding useful information in the collection. Recommender systems attempt to extract such useful information by capturing and mining one or more measures of the usefulness of the data. This paper considers the architecture of recommender systems in general, and offers a specific example called SurfAdvisor to illustrate how some design issues inherent in such systems can be addressed. Specifically, our system recommends useful documents to World Wide Web surfers. It uses a series of asynchronous processes managed by an intelligent proxy to (a) classify each Web page, (b) allow users to vote on the usefulness of the page, and (c) provide four types of recommendation on any of 1167 subject areas. In addition, our system also provides users with title, abstract, voting, and speed information about each link on a Web page, all in real time. In the end, we show that appropriate mining of data by a well-designed recommender system can significantly improve the perceived usefulness of the Web."