We are now seeing exhaustive research in a field that, until recently, did not gain a lot of attention from the academic or business communities. Academically, recommender systems present plenty of challenges for researchers from mathematics, statistics, computer science, and artificial intelligence; from the business perspective, good recommendations can lead to up-selling or cross-selling opportunities in the new economy.
This paper opens with an excellent introduction, in which the authors state their perspective, based on the classification of recommender systems and the way each system models the target user. Section 2 presents a historical summary, from information retrieval in the 1960s and 1970s, to the advanced recommender systems of today, embedded in huge e-commerce systems such as Amazon.com. Section 3 presents recommender systems that use explicit user modeling, meaning the model is obtained from surveys, explicit feedback, reviews, or keywords used in search. Section 4 is devoted to systems that use implicit modeling; the model in this case is obtained from data analysis techniques, such as data mining or knowledge discovery, applied to databases that store traces of the user behavior, as transaction logs, emails, documents retrieved, and so on. Issues regarding social networks, user targeting, and privacy are briefly addressed in section 5. Finally, section 6 outlines open problems and research opportunities in this field.
This paper is a good starting point for people who want to understand the concepts, facilities, and potential of recommender systems, and for researchers considering the topic. It includes a comprehensive list of references (a feature of a good survey), good illustrations, and a very coherent outline.