Chapter

Recommender Systems Handbook

pp 545-567

People-to-People Reciprocal Recommenders

  • Irena KoprinskaAffiliated withSchool of Information Technologies, University of Sydney Email author 
  • , Kalina YacefAffiliated withSchool of Information Technologies, University of Sydney

* Final gross prices may vary according to local VAT.

Get Access

Abstract

People-to-people reciprocal recommenders are an emerging class of recommender systems. They differ from traditional items-to-people recommenders as they must satisfy the preferences and needs of the two parties involved in the recommendation. In contrast, traditional items-to-people recommenders are one-sided and must satisfy only the preference of the person for whom the recommendation is generated. We review the characteristics and present an overview of existing reciprocal recommenders. To illustrate the various aspects of these recommenders and how reciprocity can be taken into account in building and evaluating such recommenders, we present a case study in online dating. We describe our reciprocal recommender algorithm that combines content-based and collaborative filtering and uses data from both user profiles and user interactions. We also study the differences between the implicit and explicit user preferences and show that implicit preferences, learned from user interactions, are better predictors of successful interactions. We conclude by outlining some future research directions.