Recommending books of revealed and latent interests in e-commerce
Yasuo Hirooka, Takao Terano, Yukichi Otsuka · 2002
We describe TwinFinder: a recommender system for an on-line bookstore. For developing TwinFinder, we extend the capability of conventional content-based recommendation by two novel methods: the Order-Matching Method (OMM) and the Cross-Matching Method (CMM). OMM and CMM aim to provide information of customers' revealed and latent interest, respectively. Thus, TwinFinder is able to discover new chances to sell books by CMM, while by OMM the system recommends books on customers' revealed interests. We have implemented and validated TwinFinder in the e-business system of a bookstore in Japan.