Abstract

Unifying Collaborative and Content-Based Filtering
Justin Basilico - Brown University Thomas Hofmann - Brown University
Collaborative and content-based filtering are two paradigms that have beenapplied in the context of recommender systems and user preference prediction.This paper proposes a novel, unified approach that systematically integratesall available training information such as past user-item ratings as well asattributes of items or users to learn a prediction function. The keyingredient of our method is the design of a suitable kernel or similarityfunction between user-item pairs that allows simultaneous generalizationacross the user and item dimensions. We propose an on-line algorithm (JRank)that generalizes perceptron learning. Experimental results on the EachMoviedata set show significant improvements over standard approaches.

Unifying Collaborative and Content-Based Filtering

Justin Basilico - Brown University
Thomas Hofmann - Brown University

Collaborative and content-based filtering are two paradigms that have beenapplied in the context of recommender systems and user preference prediction.This paper proposes a novel, unified approach that systematically integratesall available training information such as past user-item ratings as well asattributes of items or users to learn a prediction function. The keyingredient of our method is the design of a suitable kernel or similarityfunction between user-item pairs that allows simultaneous generalizationacross the user and item dimensions. We propose an on-line algorithm (JRank)that generalizes perceptron learning. Experimental results on the EachMoviedata set show significant improvements over standard approaches.