A Study: Hate Speech and Offensive Language Detection in Textual Data by Using RNN, CNN, LSTM and BERT Model
Anuj Kumar · 2022 6th International Conference on Intelligent Computing and Control Systems (ICICCS) · 2022
The issue of offensive speech on social networking platforms is widespread, or like face book and twitter website is dealing with it. Several approaches for intent-based text categorization have been investigated. Each technique has advantages and disadvantages depending on the type of goal, the dimension of the facts collection, the highest reach of content, and so on. Various ways for detecting dislike and incitement to hatred have been published in the literature. The primary objective of paper is to give a approximate investigation of several method for detecting hateful content and abusive content. Recurrent neural networks (RNN), convolutional neural networks (CNN), long short-term memory (LSTM), and bidirectional encoder representations from transformers are among the approaches used (BERT). The impact of class weighting methods on the efficacy of deep learning techniques was investigated. Our research shows that the or before BERT model exceed another method including both un weighted and weighted offensive prose allocation. In terms of foul language categorization, the RNN and CNN models beat every alternate methods in both un weighted and weighted cases. It was discovered that the class weighting strategy significantly improved the grading precision of all four hate speech models.