Browser-oriented universal cross-site recommendation and explanation based on user browsing logs

Abstract

Our research aims to bridge the gap between different websites to provide cross-site recommendations based on browsers. Recent advances have made recommender systems essential to various online applications, such as e-commerce, social networks, and review service websites. However, practical systems mainly focus on recommending inner-site homogeneous items. For example, a movie review website usually recommends other movies within the site when a user has enjoyed a movie online. However, it would be exciting if the system recommends some attractive products related to this movie from some e-commerce websites like Amazon or eBay.Such an ability to provide heterogeneous cross-site recommendations may shed light on brand new and promising business models, which could benefit both the online shops in expanding the marketing efforts, and the online users in discovering items of interest from a wider scope. In this research, we propose and formalize the problem of universal recommendation, record and analyze user browsing actions in web browsers, and provide browser-oriented cross-site recommendations when the users are surfing online.