Felix H. del-Olmo, Eduardo H. Martin, Elena Gaudioso
Comparing recommender systems requires common means for evaluating them. Nevertheless they have been so far evaluated through many, often incompatible, ways. This problem is mainly due to the lack of a common framework for recommender systems. In this paper, we propose such a framework. Within this framework, recommender systems are applications with the following composed objective: (i) to choose which (of the items) are going to be shown to the user and, (ii) to decide both when and how the recommendations will be shown. Starting from this framework, a natural metric emerges. We will complete the framework proposal by studying the proporties of this metric, along with a comparison of it with the traditional ones.
Subjects: 6.3 User Interfaces; 2. Architectures
Submitted: Jun 7, 2007