Recommenders for commerce, content, and community

摘要

Recommender systems are ubiquitous on the Internet for helping sell products - everything from automobiles to zebras (stuffed, anyway). Novel applications are emerging that use recommenders for non-Internet applications and that apply them to the problems of distributing content on the Internet and to developing online communities. Community-building is proving one of the most successful ways to create "stickiness" among customers. A vibrant community of practice around a company's products creates a powerful barrier to competition and enables consumers to help sell and support your products. We briefly survey eight principles of recommender systems, illuminated by examples from research and commerce. We use the principles to investigate the algorithms that underlie recommender systems, the interfaces for presenting the recommendations, the best practices for deploying them - and the easiest ways to get a recommender system badly wrong. Along the way, we consider issues of how to build a recommender community from scratch, group recommendations, and consumer privacy. We conclude with a look at some of the most important active research areas in recommender systems.