ISTS: Implicit social trust and sentiment based approach to recommender systems

Abstract

We propose a novel personalized Recommender System (RS) framework, so-called Implicit Social Trust and Sentiment (ISTS) based RS which draws user preferences by exploring the user’s Online Social Networks (OSNs). This approach overcomes the overlooked use of OSNs in Recommender Systems (RSs) and utilizes the widely available information from such networks. Bearing in mind that a user’s selection is greatly influenced by his/her trusted friends and their opinions, this paper presents a framework to apply a new source of data to personalise recommendations by mining their friends’ short text posts in microbloggings. ISTS maps suggested recommendations into numerical rating scales by applying three main components: (1) measuring the implicit trust between friends based on the intercommunication activities; (2) inferring the sentiment rating to reflect the knowledge behind friends’ short posts, so-called micro-reviews, using sentiment techniques adding several ONSs language features to empower the extracted sentiment; (3) identifying the impact degree of trust level between friends and sentiment rating from micro-reviews on recommendations by using machine learning regression algorithms including linear regression, random forest and support vector regression (SVR). Our framework takes into consideration the semantic relationships between rating categories when estimating ratings to users. Empirical results, using real social data from Twitter microblogger, verified the effectiveness and promises of ISTS.