Adapting Spreading Activation Techniques towards a New Approach to Content-Based Recommender Systems

Abstract:

Recommender systems fight information overload by selecting automatically items that match the personal preferences of each user. Content-based rec-ommenders suggest items similar to those the user liked in the past by resorting to syntactic matching techniques, which leads to overspecialized recommendations. The so-called collaborative approaches fight this problem by considering the preferences of other users, which results in new limitations. In this paper, we avoid the intrinsic downsides of collaborative solutions and diversify the content-based recommendationsby reasoning about the semantics of the user's preferences. Specifically, we present a novel domain-independent content-based recommendation strategy that exploits Spreading Activation techniques as the reasoning mechanism. Our contribution consists of adapting and extending the internals of traditional SA techniques in order to fulfill the personalization requirements of a recommender system. The resulting reasoning-driven strategy enables to discover additional knowledge about the user's preferences and leads to more accurate and diverse content-based recommendations. Our approach has been preliminary validated with a set of viewers who received recommendations of Digital TV contents. © Springer-Verlag Berlin Heidelberg 2010.

Año de publicación:

2010

Keywords:

Fuente:

scopus