Short Review of Sentiment-Based Recommender Systems

Abstract

Sentiment analysis is a trendy domain of Machine Learning which has developed considerably in the last several years. It helps to determine the sentiment of a user in an utterance, a document or a review. Some systems can extract the target of the sentiment, in order to distinguish separated sentiments over the different aspects of the product. Recommender systems (RS) are also more and more used in everyday life due to the rise of the Big Data era. We can count 3 types of Recommender Systems: the ones using Collaborative Filtering, the ones which are Content Based and the hybrid ones which are melting several kinds of information in various proportions. The general recommender systems are using global features, modeling the interest of the user on a specific topic, but they use neither the sentimental information nor the interest and preferences of the user over the different aspects that can be found in the recommended items. The opinions contained in the reviews can help to disentangle the user's preferences over the different aspects, modeling the user more confidently as well as the general opinion of the crowd over the product. By analyzing the reviews available in the Web, Sentiment Analysis systems can help improving the recommender systems wether they are simple, aspect-based or end-to-end deep models. This paper outlines a short review of the recommender systems enhanced by sentiment analysis modules.