CoRoSeOf - An Annotated Corpus of Romanian Sexist and Offensive Tweets

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 2269–2281 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 2269 CoRoSeOf - An Annotated Corpus of Romanian Sexist and Offensive Tweets Diana Constantina H¨ofels, C¸ a˘grı C¸ ¨oltekin, Irina Diana M˘adroane University of T¨ubingen, University of T¨ubingen, West University of Timis,oara diana-constantina.hoefels@student.uni-tuebingen.de, ccoltekin@sfs.uni-tuebingen.de, irina.madroane@e-uvt.ro Abstract This paper introduces CoRoSeOf, a large corpus of Romanian social media manually annotated for sexist and offensive language. We describe the annotation process of the corpus, provide initial analyses, and baseline classification results for sexism detection on this data set. The resulting corpus contains 39 245 tweets, annotated by multiple annotators (with an agreement rate of Fleiss’ κ = is 0.45), following the sexist label set of a recent study. The automatic sexism detection yields scores similar to some of the earlier studies (macro averaged F1 score of 83.07 % on binary classification task). We release the corpus with a permissive license. Keywords: corpus annotation, sexist language, offensive language, Romanian corpora 1. Introduction Sexism is defined as a “prejudice or discrimination based on sex or gender, especially against women and girls”, and dates back from the “second-wave” femi- nism.1 The term was first used by Pauline Leet at an all- male lecture at a conservative college in 1965, where she talked about how women are positioned, marginal- ized, and considered invisible (Shapiro, 1985). Study after study shows how women are vulnerable to sex- ism: biases asserting that girls are more sensitive to pain and more reactive to it than boys (Cohen et al., 2014), a culture of sexism in the academia through the under-representation of women (Savigny, 2014), women at the workplace (Barnett, 2005; Verniers and Vala, 2018), and even in Natural Language Processing (NLP) models (Sun et al., 2019; Feine et al., 2020; Bhaskaran and Bhallamudi, 2019; Kiritchenko and Mohammad, 2018). Furthermore, a number of activists have used online platforms to raise awareness about and challenge gender injustice in the recent past, in- cluding Laura Bates (Bates, 2014), Reni Eddo-Lodge, Tarana Burke, and Lucy-Anne Holmes. Nearly 90% of the world’s population – both men and women – have prejudices against women (United Nations Devel- opment Programme, 2020). As a prevalent aspect of the world’s societies, it comes as no surprise that sexism permeates the language do- main, with a debate among female activists dating back to the 19th century. Media studies, theory, and activism have challenged media representations of women, pro- viding a platform for exploring social norms and expec- tations (Savigny, 2020). Sexist language is also used in social-media communication, leading to an envi- ronment of intimidation, insecurity, fear, and violence. Furthermore, sexist discourse on social media may fa- cilitate the spreading, justification, and reinforcement 1https://www.britannica.com/topic/ sexism of sexist acts. Thus, in order to protect the users of these platforms from sexist language, it is imperative to recognize, detect, and prevent sexist language in the online context. In this study, we contribute to this pur- …

Content truncated to 500 words