Developing corpora and tools for sentiment analysis: the experience of the University of Turin group

The paper describes the ongoing experience at the University of Turin in developing linguistic resources and tools for sentiment analysis of social media. We describe in particular the development of Senti-TUT, a human annotated corpus of Italian Tweets including labels for sentiment polarity and irony, which has been recently exploited within the SENTIment POLarity Classification shared task at Evalita 2014. Furthermore, we report about our ongoing work on the Felicittà web-based platform for estimating happiness in Italian cities, which provides visualization techniques to interactively explore the results of sentiment analysis performed over Italian geotagged Tweets.