Speaker: Prof. Paolo Rosso, Polytechnic University of Valencia

With the increasing of social media, consumers rely more than ever on online reviews to make their decisions. A recent survey found that 87% of them have reinforced their decisions to purchase a product due to positive online reviews. At the same time 80% of consumers have changed their minds on the basis of negative information they found online. Therefore, online opinions play an important role for companies and there is an increasing trend to post fake reviews with the aim to sound authentic and deceive the consumers. The detection of deceptive opinions is a quite challenging problem and not only automatically (only 60% of humans discriminate between truthful and deceptive opinions with a certain degree of accuracy). In the first part of the talk I will give an overview of the state-of-the-art approaches for detecting deceptive opinions. The second part of the talk is devoted to another issue that makes the life of a sentiment analysis tool quite difficult: detecting irony. In ironic opinions what is literally said is usually negated, and in absence of an explicit negation marker. This makes sentiment analysis quite challenging. Therefore, there is a growing interest from the research community in investigating the impact of irony on sentiment analysis and a task has been organized recently at SemEval in 2015 on sentiment analysis of figurative language in Twitter. In the talk I will describe how irony is employed in tweets and reviews and what are the recent state-of-the-art attempts for its automatic detection. Linguistic devices such as ambiguity, incongruity, unexpectedness and emotional contexts play an important role as triggers of irony. At the end I will also address the even more challenging fine-grained problem of discriminating between irony and sarcasm: e.g. If you find it hard to laugh at yourself, I would be happy to do it for you.