Abusive Language Detection from Real-Time Radio Message Gateway

Abstract

Most of the research work on identifying objectionable texts is in English and some of them can detect objectionable texts but there are some works available in the Bengali language. People in Bangladesh usually feel comfortable expressing their views in Bengali on media platforms. The identification of offensive text in the Bengali language will be effective in preventing cybercrime, which has become a major concern of Bangladesh nowadays in preventing online language harassment, blackmailing, and cyberbullying. It is also challenging to include them in different groups depending on the text. In this project, we present our work on detecting abusive language from real-time radio message gateway. Usually what happens on radio stations is that users give messages but some messages are good and some messages are offensive. They follow all messages except filtered from their server. We have created a data set of about 45,000 messages, including good messages and offensive messages. We have leveled the good and offensive messages. We have leveled the good messages as 1 and the offensive messages as 0. Our algorithm will check the dataset when a live radio program is broadcast and listeners will try to communicate by sending a direct message and will only show positive comments. In real-time, all abusive messages will come to the spam folder here to save time.