Multilingual Offensive Language Identification with Cross-lingual Embeddings

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 5838–5844, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 5838 Multilingual Offensive Language Identification with Cross-lingual Embeddings Tharindu Ranasinghe University of Wolverhampton Wolverhampton, UK tharindu.ranasinghe@wlv.ac.uk Marcos Zampieri Rochester Institute of Technology Rochester, NY, USA mazgla@rit.edu Abstract Offensive content is pervasive in social me- dia and a reason for concern to companies and government organizations. Several stud- ies have been recently published investigat- ing methods to detect the various forms of such content (e.g. hate speech, cyberbulling, and cyberaggression). The clear majority of these studies deal with English partially be- cause most annotated datasets available con- tain English data. In this paper, we take ad- vantage of English data available by applying cross-lingual contextual word embeddings and transfer learning to make predictions in lan- guages with less resources. We project pre- dictions on comparable data in Bengali, Hindi, and Spanish and we report results of 0.8415 F1 macro for Bengali, 0.8568 F1 macro for Hindi, and 0.7513 F1 macro for Spanish. Fi- nally, we show that our approach compares fa- vorably to the best systems submitted to recent shared tasks on these three languages, confirm- ing the robustness of cross-lingual contextual embeddings and transfer learning for this task. 1 Introduction Offensive posts on social media result in a num- ber of undesired consequences to users. They have been investigated as triggers of suicide attempts and ideation, and mental health problems (Bonanno and Hymel, 2013; Bannink et al., 2014). One of the most common ways to cope with offensive content online is training systems to be capable of recogniz- ing offensive messages or posts. Once recognized, such offensive content can be set aside for human moderation or deleted from the respective platform (e.g. Facebook, Twitter), preventing harm to users and controlling the spread of abusive behavior in social media. There have been several recent studies published on automatically identifying various kinds of of- fensive content such as abuse (Mubarak et al., 2017), aggression (Kumar et al., 2018, 2020), cyber-bullying (Rosa et al., 2019), and hate speech (Malmasi and Zampieri, 2018). While there are a few studies published on languages such as Arabic (Mubarak et al., 2020) and Greek (Pitenis et al., 2020), most studies and datasets created so far in- clude English data. Data augmentation (Ghadery and Moens, 2020) and multilingual word embed- dings (Pamungkas and Patti, 2019) have been ap- plied to take advantage of existing English re- sources to improve the performance in systems dealing with languages other than English. To the best of our knowledge, however, state-of-the-art cross-lingual contextual embeddings such as XLM- R (Conneau et al., 2019) have not yet been applied to offensive language identification. To address this gap, we evaluate the performance of cross-lingual contextual embeddings and transfer learning (TL) methods in projecting predictions from English to other languages. We show that our methods com- pare favorably to state-of-the-art approaches sub- mitted to recent shared tasks on all datasets. The main contributions of …

Content truncated to 500 words