Research of techniques used in toxicity detection

• Main Authors: Kumar, Aakash, Chauhan, Aarjav, Babbar, Abhishek, Kaur, Gull
• Format: Conference Proceeding
• Language: English
• Subjects: Classification; Machine learning
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0148415

In this world of ever-increasing texts, chats and literature users nowadays are posting countless comments on various social media platforms, news websites, and discussion forums. While some comments bring the positive side of the matter, whileother remarks are detrimental or combative in nature. Manually monitoring of such a large volume of comments and chats is a tedious and a time taking task. Therefore, various machine learning based techniques are employed to cater the task of detecting the levels of toxicity in the comments. We gathered information from 20 primary research papers that were found to be relevant in the field of toxic comment classification. We explored all the papers thoroughly taking into consideration the publishing date andconferences in which they were published. We noted the data set used, performance criteria, machine learning techniques employed and other important features such as diversity of toxicity classes and comparative report stated in them. We concluded our work with a detailed overview of the existing issues and a suggestion list of future topics that could be explored and implemented to make the system more accurate in classifying the toxic comments.