JCTICOL at SemEval-2019 Task 6: Classifying Offensive Language in Social Media using Deep Learning Methods, Word/Character N-gram Features, and Preprocessing Methods

Proceedings of the 13th International Workshop on Semantic Evaluation (SemEval-2019), pages 645–651 Minneapolis, Minnesota, USA, June 6–7, 2019. ©2019 Association for Computational Linguistics 645 Abstract In this paper, we describe our submissions to SemEval-2019 task 6 contest. We tackled all three sub-tasks in this task “OffensEval - Identifying and Categorizing Offensive Language in Social Media”. In our system called JCTICOL (Jerusalem College of Technology Identifies and Categorizes Offensive Language), we applied various supervised ML methods. We applied various combinations of word/character n- gram features using the TF-IDF scheme. In addition, we applied various combinations of seven basic preprocessing methods. Our best submission, an RNN model was ranked at the 25th position out of 65 submissions for the most complex sub-task (C). 1 Introduction Offensive language is frequent in social media. For instance, ScanSafe's monthly "Global Threat Report" reported that up to 80% of blogs contained offensive contents and 74% included porn in the format of the image, video, or offensive language (Cheng, 2007). There are people that take advantage of the perceived anonymity of computer-mediated communication, using this to write in behavior that many of them would not consider in real life. Online news and social networking services, online communities, social media platforms, and various computer companies have been investing a lot of effort, time and money to cope with offensive language in order to prevent abusive behavior. Computational methods are among the most effective strategies to identify various types of aggression, offense, and hate speech in user- generated content (e.g., comments, microblogs, posts, and tweets). Detection of offensive language has been investigated in recent years in various studies (Waseem et al. 2017; Davidson et al., 2017, Malmasi and Zampieri, 2018, Kumar et al. 2018) and various workshops such as ALW (Abusive Language Online) and TRAC (Trolling, Aggression, and Cyberbullying). In this paper, we describe our submissions to SemEval-2019 task 6 contest. In Task-6, OffensEval, there are three different sub-tasks. Sub-task A deals with offensive language identification. Sub-task B deals with the automatic categorization of offense types. Sub- task C deals with offense target identification. The report of the OffensEval task is described in Zampieri et al. (2019A) and the description of the OLID dataset that was used for the competition is in Zampieri et al. (2019B). The structure of the rest of the paper is as follows. Section 2 discusses work related to offensive language in social media, tweet classification, and data preprocessing. Section 3 presents, in general, the task description. In Section 4, we describe the submitted models and their experimental results. Section 6 summarizes and suggests ideas for future research. 2 Background 2.1 Offensive Language in Social Media In recent years, there has been an increase in the number of studies dealing with Offensive language in social media. Nobata et al. (2016) developed a machine learning based method to detect hate speech on online user comments from two domains. They also built a corpus of user comments annotated accordingly to three subcategories (hate speech, derogatory, JCTICOL at …

Content truncated to 500 words