Sentiment detection in social networks and in collaborative learning environments

Daily millions of messages appear on the web, which is becoming a rich source of data for opinion mining and sentiment analysis. The computational study of opinions, feelings and emotions expressed in a text often relates to the identification of agreement or disagreement with statements, contained in comments or reviews, that convey positive or negative feelings. The detection and analysis of sentiment in textual communication is a topic attracting attention also in the context of collaborative learning in social networks, being learners actively engaged in presenting and defending ideas and opinions, as well as exchanging moods about courses with peers. In this paper, we investigate the adoption of a probabilistic approach based on the Latent Dirichlet Allocation (LDA) as Sentiment Grabber. Through this approach, for a set of documents belonging to a same knowledge domain, a graph, the Mixed Graph of Terms, can be automatically extracted. The paper shows how this graph contains a set of weighted word pairs, which are discriminative for sentiment classification. The proposed method has been tested in different context: a standard dataset containing movie reviews; a real-time analysis of social networks posts; a collaborative learning scenario. The experimental evaluation shows how the proposed approach is effective and satisfactory.