Invited Talk

Sara Tonelli, Fondazione Bruno Kessler, Trento, Italy

Sara Tonelli holds a PhD in Language Sciences from Università Ca’ Foscari, Venice. Since 2013 she has been the head of the Digital Humanities research group at Fondazione Bruno Kessler in Trento, Italy. Among many projects in digital humanities, she is currently involved in the H2020 ODEUROPA project (focusing on olfactory information extraction), and in the H2020 PERCEPTIONS project (online perception and migration narratives related to EU). Since January 2021 she is also the scientific coordinator of the KID ACTIONS European project (addressing cyberbullying among children and adolescents). Sara’s main research interests are related to temporal and event-based processing of texts, especially in the historical domain, and social media processing, including the detection of abusive language.

Dissecting offensive language detection: does it work, and what can we do with it?

Social media messages are often written to attack specific groups of users based on their religion, ethnicity or social status, and they can be particularly threatening to vulnerable users such as teenagers. It is therefore very important to develop reliable, unbiased and robust detection systems to support stakeholders in fighting online hatred. Although state-of-the-art systems yield very good classification results, the problem is far from being solved. In my talk, I will discuss which issues still affect the development of abusive language detection systems, for example the problem of dealing with annotators’ disagreement in the creation of training data, and the issues related to contextual information in threads. On the other hand, I will show how the output of offensive language detection systems can be integrated with network-based information to study the behavioral patterns of different types of users, also in relation to misinformation.