CiTIUS-COLE at SemEval-2019 Task 5: Combining Linguistic Features to Identify Hate Speech Against Immigrants and Women on Multilingual Tweets

Proceedings of the 13th International Workshop on Semantic Evaluation (SemEval-2019), pages 387–390 Minneapolis, Minnesota, USA, June 6–7, 2019. ©2019 Association for Computational Linguistics 387 CiTIUS-COLE at SemEval-2019 Task 5: Combining Linguistic Features to Identify Hate Speech Against Immigrants and Women on Multilingual Tweets Sattam Almatarneh (CiTIUS) Universidade de Santiago de Compostela, Spain University of Vigo, Spain sattam.almatarneh@usc.es Pablo Gamallo (CiTIUS) Universidade de Santiago de Compostela, Spain pablo.gamallo@usc.es Francisco J. Ribadas Pena Department of Computer Science University of Vigo, Spain ribadas@uvigo.es Abstract This article describes the strategy submitted by the CiTIUS-COLE team to SemEval 2019 Task 5, a task which consists of binary clas- sification where the system predicts whether a tweet in English or in Spanish is hateful against women or immigrants or not. The proposed strategy relies on combining linguis- tic features to improve the classifier’s perfor- mance. More precisely, the method combines textual and lexical features, embedding words with the bag of words in Term Frequency- Inverse Document Frequency (TF-IDF) repre- sentation. The system performance reaches about 81% F1 when it is applied to the training dataset, but its F1 drops to 36% on the official test dataset for the English and 64% for the Spanish language concerning the hate speech class. 1 Introduction Hate speech is usually defined as any communica- tion that derogates a person or a group based on some characteristic such as race, color, ethnicity, gender, sexual orientation, nationality, religion, or another characteristic (Schmidt and Wiegand, 2017). The spread of the Internet and the increas- ing use of social networks has led people to have an increased willingness to express their opinions online. Despite the great benefits of using the In- ternet and particularly the social networks, the risk is that people are more likely to adopt aggressive behavior because of the anonymity provided by these environments. This contributes to the prop- agation of hate speech as well. Since this type of tendentious communication can be extremely harmful to society, governments and social net- work platforms can benefit from detection and pre- vention tools. The scientific study of hate speech, from a computer science point of view, is recent, and the number of studies in the field is low (For- tuna and Nunes, 2018). The goal of SemEval- 2019 Task 5 as described in Basile et al. (2019) is Hate Speech detection in Twitter focused on two specific targets, women and immigrants. The task is organized in two related sub-tasks for each lan- guage (English and Spanish): • TASK A - Hate Speech Detection against Im- migrants and Women • TASK B - Aggressive behavior and Target Classification. In this article, we describe our proposed system for task A only. Our approach is mainly based on the generation of corpus-based dictionaries containing hate speech words which are used in addition to other linguistic features to improve the efficiency in detecting hate speech in both English and Span- ish languages. This paper is organized as follows. The method is described in Section 2. Experiments, results, …

Content truncated to 500 words