Influence of the inclusion of unstructured data in recommender systems

Abstract

The vast amount of data available on the web makes the need for systems that help us to distinguish relevant content from irrelevant content more than evident. This task is carried out by the so-called Recommender Systems, and we can currently find them on most of the websites we use on a daily basis. These systems typically learn from the user’s consumption history (purchases, listens, clicks, ...) but rarely make use of additional information provided by users in the form of natural language texts or images (unstructured data). In the multiple works of this thesis we intend to take advantage of this type of information in different ways with the aim of improving the performance of Recommender Systems as well as users experience, the ways in which to obtain a recommendation or the presentation of the final recommendations among others.