Well-Argued Recommendation: Adaptive Models Based on Words in Recommender Systems

Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 1943–1947, Seattle, Washington, USA, 18-21 October 2013. c⃝2013 Association for Computational Linguistics Well-argued recommendation: adaptive models based on words in recommender systems Julien Gaillard University of Avignon Agorantic Avignon, France julien.gaillard@univ-avignon.fr Marc El-Beze University of Avignon Agorantic Avignon, France marc.elbeze@univ-avignon.fr Eitan Altman INRIA Sophia Antipolis Agorantic Sophia-Antipolis, France eitan.altman@inria.fr Emmanuel Ethis University of Avignon Agorantic Avignon, France emmanuel.ethis@univ-avignon.fr Abstract Recommendation systems (RS) take advan- tage of products and users information in order to propose items to consumers. Collaborative, content-based and a few hybrid RS have been developed in the past. In contrast, we propose a new domain-independent semantic RS. By providing textually well-argued recommenda- tions, we aim to give more responsibility to the end user in his decision. The system includes a new similarity measure keeping up both the accuracy of rating predictions and coverage. We propose an innovative way to apply a fast adaptation scheme at a semantic level, provid- ing recommendations and arguments in phase with the very recent past. We have performed several experiments on films data, providing textually well-argued recommendations. 1 Introduction Recommender systems aim at suggesting appropri- ate items to users from a large catalog of products. Those systems are individually adapted by using a specific profile for each user and item, derived from the analysis of past ratings. The last decade has shown a historical change in the way we consume products. People are getting used to receive recom- mendations. Nevertheless, after a few bad recom- mendations, users will not be convinced anymore by the RS. Moreover, if these suggestions come without explanations, why people should trust it? Numbers and figures cannot talk to people. To answer these key issues, we have designed a new semantic recommender sytem (SRS) including at least two innovative features: • Argumentation: each recommendation relies on and comes along with a textual argumenta- tion, providing the reasons that led to that rec- ommendation. • Fast adaptation: the system is updated in a con- tinuous way, as each new review is posted. In doing so, the system will be perceived as less intrusive thanks to well-chosen words and its fail- ures will be smoothed over. It is therefore necessary to design a new generation of RS providing textu- ally well-argued recommendations. This way, the end user will have more elements to make a well- informed choice. Moreover, the system parameters have to be dynamically and continuously updated, in order to provide recommendations and arguments in phase with the very recent past. To do so, we have adapted the algorithms we described in Gail- lard (Gaillard et al., 2013), by including a semantic level, i.e words, terms and phrases as they are natu- rally expressed in reviews. This paper is structured as follows. In the next section, we present the state of the art in recom- mendation systems and introduce some of the im- provements we have made. Then, we present our approach and define the associated methods …

Content truncated to 500 words