Comparative analysis of the performance of Machine Learning and Transfer Learning models in detecting hate on Twitter
Shreyansh Khandelwal, Aruna M G · 2022 2nd International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE) · 2022
As more and more individuals hop on the social media train, there is an exponential increase in hate speech on these platforms, especially Twitter, where there is no moderation present and users can post even Not Safe For Work (NSFW) content. Moderation of content is an impossible task if done manually. One cannot look at all the content and verify if it's safe or not. So, to tackle this problem, one needs the aid of Machine Learning and Transfer Learning. Hence, we have developed and compared three different Machine Learning models in this paper, namely: Logistic Regression model, Simple Neural Network model, and a Bidirectional Encoder Representations from Transformers (BERT) model trained on the same dataset, which would classify the given sentence as a “hate comment” or “not a hate comment,” to ensure a safe environment for all users on the platform.