Trust and user profiling for refining the prediction of reader's emotional state induced by news articles

Abstract

Automatic evaluation is an efficient alternative of capturing the sentiments and the attitude of a targeted audience towards a specific topic or subject. The starting context of our research is represented by the social media's important role in everybody's life. As the social media includes the web technologies that enable us to communicate directly and to modify user-generated content, the adoption of such online communication channels, as well as social networks (e.g., Facebook, Twitter, Google+) or Computer Supported Collaborative Learning (CSCL) technologies (e.g., chat, wiki, blog, forum) have gained an increasing trend and have reshaped interaction and information access. The purpose of this paper is to present an overview of opinion mining techniques, to describe the implementation of a previously developed system within our research group - Emotion Monitor -, alongside with our current improvements, such as the new trust module for evaluating the system's confidence in the current user, as well as enriching the user's profile in order to further personalize the generated results. In the end, the system predicts the manner in which a news article is affecting the emotional state of a user by integrating specific natural language processing techniques (especially Latent Semantic Analysis) and the reader's profile.