Regresar

Explainable Recommender Systems: From Theory to Practice

Abstract:

Recommender systems support users’ decision-making, and they are key for helping them discover resources or relevant items in an information-overloaded environment such as the web. Like other Artificial Intelligence-based applications, these systems suffer from the problem of lack of interpretability and explanation of their results. Enriching or augmenting the system output with explanations increases the users’ trustworthiness and reliability regarding the system decisions. Therefore, it is important not only to measure the performance of automatic models but also to measure the explainability of the system. In this paper, we present research related to explainable recommender systems and a demonstrative case. To illustrate how explainable recommendations can be generated, we present two scenarios based on the Tripadvisor dataset.