SINAI at SemEval-2019 Task 5: Ensemble learning to detect hate speech against inmigrants and women in English and Spanish tweets

Proceedings of the 13th International Workshop on Semantic Evaluation (SemEval-2019), pages 476–479 Minneapolis, Minnesota, USA, June 6–7, 2019. ©2019 Association for Computational Linguistics 476 SINAI at SemEval-2019 Task 5: Ensemble learning to detect hate speech against inmigrants and women in English and Spanish tweets Flor Miriam Plaza-del-Arco, M. Dolores Molina-Gonz´alez, M. Teresa Mart´ın-Valdivia, L. Alfonso Ure˜na-L´opez Department of Computer Science, Advanced Studies Center in ICT (CEATIC) Universidad de Ja´en, Campus Las Lagunillas, 23071, Ja´en, Spain {fmplaza, mdmolina, maite, laurena}@ujaen.es Abstract Misogyny and xenophobia are some of the most important social problems. With the in- crease in the use of social media, this feeling of hatred towards women and immigrants can be more easily expressed, therefore it can cause harmful effects on social media users. For this reason, it is important to develop systems ca- pable of detecting hateful comments automat- ically. In this paper, we describe our system to analyze the hate speech in English and Spanish tweets against Immigrants and Women as part of our participation in SemEval-2019 Task 5: hatEval. Our main contribution is the integra- tion of three individual algorithms of predic- tion in a model based on Vote ensemble classi- fier. 1 Introduction With the growing prominence of social media like Twitter or Facebook, more and more users are publishing content and sharing their opinions with others. Unfortunately, the content often contains hate speech language that can have damaging ef- fects on social media users. This fact concerns to social media platforms like Facebook since ac- cording to an EU’s report, it removes 82 percent of illegal hate speech on the platform, up from 28 percent in 20161. Normally, hate speech can be aimed at a per- son or a group base on some characteristic such as race, sexuality, color, ethnicity, physical ap- pearance, religion, among others (Erjavec and Kovaˇciˇc, 2012). Currently, two of the targets most affected by these types of offensive com- ments are immigrants and women (Waseem and Hovy, 2016). In particular, when the hate speech is gender-oriented, and it specifically targets women, we refer to it as misogyny (Manne, 2017) and 1https://cnb.cx/2RGmEwel when the hate speech is against immigrants, we refer to it as xenophobia (Sanguinetti et al., 2018). Recently, a growing number of researchers have started to focus on studying the task of automatic detection of hateful language online (Fortuna and Nunes, 2018; Fersini et al., 2018b), moreover, some academic events and shared tasks have taken place focusing on this issue (Fersini et al., 2018a). It is consider as a difficult task for social media platforms. For example, popular social media such as Twitter, Instagram or Facebook are not able to automatically solve this problem and depend on their community to report hateful speech content. The severe consequences of this problem, com- bined with the large amount of data that users pub- lish daily on the Web, requires the development of algorithms capable of automatically detecting in- appropriate online remarks. In this paper, we describe our participation in SemEval-2019 Task 5: …

Content truncated to 500 words