Analysis of Hate Tweets Using CBOW-based Optimization Word Embedding Methods Using Deep Neural Networks

Abstract

In the recent years, governments, corporations, and academics have poured money into addressing the rising frequency of hate speech and expression, as well as the pressing need for effective solutions. Many methods for identifying hate speech have been developed and published on the Internet. This seeks to categorise textual information as non-hate or hate speech, with the algorithm being able to detect targeted features in the latter instance. which attempts to detect if textual content is non-hate or hate speech, with the algorithm recognising the desired qualities in the latter case Our proposed study uses Continuous Bag Of Word (CBOW)-based word embedding to try to predict the target term by analysing the context of the surrounding words, and feature extractors, deep learning-based structures explicitly capture the meanings of offensive speech. On the largest collection of hate speech data sets based on Twitter, our approaches are sorely tested. For hate speech identification, we investigate the influence of several extra-linguistic factors in combination with character n-grams. In addition, we provide a lexicon based on the most important words in our data. The proposed approach predicts 95% word embedding accuracy in real-time Twitter hate speech test data.