Sentiment analysis in social networks: A methodology based on the latent dirichlet allocation approach

The detection and analysis of sentiment in textual communication is a topic attracting attention in both academia and industry. In fact, thanks to the explosion of the Social Networks a wealth of information is produced every day. This huge amount of contents can be very helpful in assessing the general public's sentiment and opinions toward products, services and topics. This paper presents a methodology for the detection of sentiment in textual contents using a methodology based on the Latent Dirichlet Allocation (LDA) approach and a word-based graphical model, the mixed graph of terms. The method has been tested in various operative scenarios: on standard datasets, on datasets obtained collecting tweets from twitter and on datasets coming from social networks as Twitter and TripAdvisor. The experimental campaigns show that the proposed approach is effective and furnishes good and reliable results in each context.