The Titans at SemEval-2019 Task 6: Offensive Language Identification, Categorization and Target Identification

Proceedings of the 13th International Workshop on Semantic Evaluation (SemEval-2019), pages 759–762 Minneapolis, Minnesota, USA, June 6–7, 2019. ©2019 Association for Computational Linguistics 759 The Titans at SemEval-2019 Task 6: Offensive Language Identification, Categorization and Target Identification Avishek Garain Computer Science and Engineering Jadavpur University, Kolkata avishekgarain@gmail.com Arpan Basu Computer Science and Engineering Jadavpur University, Kolkata arpan0123@gmail.com Abstract This system paper is a description of the sys- tem submitted to “SemEval-2019 Task 6”, where we had to detect offensive language in Twitter. There were two specific target audi- ences, immigrants and women. The language of the tweets was English. We were required to first detect whether a tweet contains offensive content, and then we had to find out whether the tweet was targeted against some individ- ual, group or other entity. Finally we were re- quired to classify the targeted audience. 1 Introduction Offensive language is pervasive in social media. Individuals frequently take advantage of the per- ceived anonymity of computer-mediated commu- nication, using this to engage in behavior that many of them would not consider in real life. Online communities, social media platforms, and technology companies have been investing heavily in ways to cope with offensive language to prevent abusive behavior in social media. One of the most effective strategies for tack- ling this problem is to use computational meth- ods to identify offense in user-generated content (e.g. posts, comments, microblogs, etc.). This topic has attracted significant attention in recent years of various Natural Language analysts. The SemEval 2019 task 6 (Zampieri et al., 2019b) was a classification task where we were re- quired to classify a tweet, as hate speech or other- wise. However, there were some additional chal- lenges presented, which involved automatic cate- gorization of offense target types and the specific detection of the target audience, namely, women or immigrants. The task was divided into three parts. In the first subtask our system categorized the instances into OFF and NOT. In the second subtask our sys- tem categorized instances into TIN and UNT while in the third subtask systems should categorize in- stances into IND, GRP, and OTH. To solve the task in hand we built a bidirectional LSTM based neural network for prediction of the classes present in the provided dataset. The paper has been organized as follows. Sec- tion 2 describes a brief survey on the relevant work done in this field. Section 3 describes the data, on which, the task was performed. The method- ology followed is described in Section 4. This is followed by the results and concluding remarks in Section 5 and 6 respectively. 2 Related Work Papers published in the last two years include the surveys by Schmidt and Wiegand (2017) and For- tuna and Nunes (2018). The paper by Davidson et al. (2017) presenting the Hate Speech Detec- tion dataset were used in (Malmasi and Zampieri, 2017) and a few other recent papers such as (ElSh- erief et al., 2018; Gamb¨ack and Sikdar, 2017; Zhang et al., 2018). A …

Content truncated to 500 words