PAN at CLEF 2015

To detect deception in communications is a difficult task for humans and a critical issue in police investigations. In fact, no specific signs of deception, such as the Pinocchio's growing nose, have never been clearly identified, even though several approaches have been developed in order to unmask liars and the false information they convey. The speech will examine the problem in the perspective of police practices, from collection to evaluation of testimonies. The contribution of different techniques and technologies for testimonies' analysis will be discussed, with particular focus on the role of the modern stylometry, as many studies in literature suggest that the discipline, which exploits computational methods in order to analyze samples of spoken and written language through their stylistic features, can be effectively employed in deception detection.

In this talk I will describe recent research within the CLiPS research centre on author profiling: the automatic assignment of demographic and psychological properties to (unknown) authors of text on the basis of linguistic analysis of these texts. I will describe different ways in which the results of this research are currently being applied. In the AMiCA project, the goal is to help moderators of social networks to detect harmful situations in their network. Our case studies concern cyberbullying, pedophile grooming, and suicide announcements. I wil show how profiling information can help achieve these tasks. In addition I will briefly demo the profiling system of Textgain, a spin-off company from CLiPS, and describe some of the applications in which their profiling web services are put to use.