Automatic discovery of abusive thai language usages in social networks

• Main Authors: Suppawong Tuarob, Jarernsri L. Mitrpanont
• Other Authors: Mahidol University
• Format: Conference or Workshop Item
• Published: 2018
• Subjects: Computer Science
• Online Access: https://repository.li.mahidol.ac.th/handle/123456789/42341

© 2017, Springer International Publishing AG. Social networks have become a standard means of communication that allows a massive amount of users to interact and consume information anywhere and anytime. In Thailand, millions of users have access to social networks, a majority of which include young children. The colloquial nature of social media inherently encourages certain expressions of language that do not conform to the standard, some of which may be considered abusive and offensive. Such ill-mannered language fashion has become increasingly used by a large number of Thai social media users. If these abusive languages are exposed to adolescents without proper guidance, they could compulsorily develop a familiar attitude towards such language styles. To address the issue, we present a set of algorithms based on machine learning, that automatically detect abusive Thai language in social networks. Our best results yield 86% f-measure (88.73% precision and 83.53% recall).