Empowering Multilingual Insensitive Language Detection: Leveraging Transformers for Code-Mixed Text Analysis

Social media usage was made easier by the internet's expanding accessibility, which also encouraged people to share their thoughts openly. However, it also gives content polluters a platform to spread objectionable postings or contents. Most of these inflammatory postings are written in many languages and are therefore simple to dodge internet monitoring programmes. This study serves as an example of our work towards the EACL 2021 joint task on the identification of offensive language in Dravidian languages. The identification of offensive language in the major social media sites was previously discovered. The necessity to identify offensive language in multilingual messages that are substantially code-mixed or written in a non-native script has arisen as a result of the increase in user diversity. The dataset consists of local languages such as Kannada, Malayalam, Tamil with English code-matching. To acquire the requirements used machine learning method (LR, SVM), deep learning technique-LSTM, and transformer-m-BERT. [RESULT The suggested models received a weighted f1 score of 0.75 (for Tamil), 0.95 (for Malayalam), and 0.71 (for Kannada).]