A BERT-based Ethical Framework for Real-Time Toxicity Detection and Moderation in Social Media

The transformer model is a crucial part of Natural Language Processing. Among all the transformers, Bidirectional Encoder Representations from Transformers, otherwise known as BERT, is one of the models that have greatly impacted the way computers are able to understand and process text information. The traditional model used to understand Natural Language is often faced with difficulties in trying to obtain a deep understanding of the information provided, as it only does so in one way. For example, in processing a sentence that goes from left to right. In the bidirectional model of BERT, it is able to process in both ways, which are right to left and left to right. BERT is primarily used in this research work for the purpose of understanding the text as well as obtaining relevant information from a large number of documents. BERT can be used for its capacity to provide a deep understanding of words, based on the context in which the words are used. The results achieved from the experiment have proved that the transformer models, like BERT, are much more superior to the rule-based systems and statistical approaches. The most important benefit that BERT provides is its efficiency in handling deeper contexts by providing information more easily and quickly as it handles deeper contexts in almost all NLP tasks.