14th Dezember 2022 17 Uhr/ 5pm Hörsaal 2/ HS 2
Erkennung von Beleidigungen mithilfe computerlinguistischer Verfahren
Prof. Dr. Michael Wiegand
In this presentation, a brief overview of the state of the art in abusive language detection will be given. One key difficulty is to build appropriate gold standards for the task which serve as
a basis for machine learning methods. In that context, the phenomenon of “spurious correlations” will be illustrated. In terms of classifiers, a lexicon-based approach will be outlined. Unlike
previous off-the-shelf methods that are usually treated as a black box, lexicon-based approaches are more explainable, less susceptible to overfitting and more stable across different domains. Predictive word lists can be compiled in a resource-intensive way combining various sources of linguistic information. As an alternative, a data-driven less resource-intensive induction method relying on emojis with an abusive connotation will also be presented.
(Please note that the presentation will be held
Michael Wiegand obtained his PhD at Saarland University in 2011. Until 2018, he had been a postdoctoral researcher at the Department for Spoken Language Systems at Saarland University. In 2019, he was research group leader in the Leibniz ScienceCampus Empirical Linguistics and Computational Language Modeling (Leibniz Institute for the German Language, Mannheim/Heidelberg University). Since 2020, he has been a professor for Computational Linguistics in the Digital Age Research Center (D!ARC) at University of Klagenfurt.