Semantics for social media

In this paper we present four experiments on the analysis Italian social media texts using a linguistically-based semantic approach. The experiments are respectively: two on newspaper articles about two political crises, one on a twitter corpus centered on political themes, and one on a case study of strategic plan programs of candidates to the presidency of our university. The analyses carried out by the same system, focus on semantic features of texts highlighting three main traits: “factivity” or factuality, “subjectivity” and polarity. The system uses semantic knowledge derived from deep linguistic analysis at propositional level to classify texts at a fine-grained level. As will be shown in the paper, linguistically-based semantic information allows for neat distinction of writing styles when comparing newspapers writing styles, for irony detection in tweets, and in different degrees, for making readability judgements.