Offensive Language and Hate Speech Detection for Danish

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 3498–3508 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 3498 Offensive Language and Hate Speech Detection for Danish Gudbjartur Ingi Sigurbergsson, Leon Derczynski IT University of Copenhagen Denmark ld@itu.dk Abstract The presence of offensive language on social media platforms and the implications this poses is becoming a major concern in modern society. Given the enormous amount of content created every day, automatic methods are required to detect and deal with this type of content. multilingual.Until now, most of the research has focused on solving the problem for the English language, while the problem is We construct a Danish dataset DKHATE containing user-generated comments from various social media platforms, and to our knowledge, the first of its kind, annotated for various types and target of offensive language. We develop four automatic classification systems, each designed to work for both the English and the Danish language. In the detection of offensive language in English, the best performing system achieves a macro averaged F1-score of 0.74, and the best performing system for Danish achieves a macro averaged F1-score of 0.70. In the detection of whether or not an offensive post is targeted, the best performing system for English achieves a macro averaged F1-score of 0.62, while the best performing system for Danish achieves a macro averaged F1-score of 0.73. Finally, in the detection of the target type in a targeted offensive post, the best performing system for English achieves a macro averaged F1-score of 0.56, and the best performing system for Danish achieves a macro averaged F1-score of 0.63. Our work for both the English and the Danish language captures the type and targets of offensive language, and present automatic methods for detecting different kinds of offensive language such as hate speech and cyberbullying. Keywords: abusive language, hate speech detection, Danish 1. Introduction Offensive language in user-generated content on online platforms and its implications has been gaining attention over the last couple of years. This interest is sparked by the fact that many of the online social media platforms have come under scrutiny on how this type of content should be detected and dealt with. It is, however, far from triv- ial to deal with this type of language directly due to the gigantic amount of user-generated content created every day. For this reason, automatic methods are required, using natural language processing (NLP) and machine learning techniques. The task of finding these poses a pressing and formidable challenge (Vidgen et al., 2019). Given the fact that the research on offensive language detection has to a large extent been focused on the En- glish language, we set out to explore the design of mod- els that can successfully be used for both English and Danish. To accomplish this, an appropriate dataset must be constructed, annotated with the guidelines described in Zampieri et al. (2019a). We furthermore set out to an- alyze relevant linguistic features by analyzing …

Content truncated to 500 words