Big language models (LLMs) perform exceptionally well in a wide range of applications, such as sentiment analysis, translation, and summarization, in addition to language production. Text classification is one fascinating use. This becomes relevant in the field of identifying poisonous or hateful communication, which is full with obstacles and moral conundrums. Our research aims to accomplish two goals. Firstly, we will provide a review of previous studies that focus on LLMs as classifiers and highlight their effectiveness in identifying and categorizing offensive or harmful information. The effectiveness of various LLMs in categorizing hate speech is then investigated. shedding light on the elements that go into an LLM’s ability – or lack thereof – to identify hateful content. Our research aims to provide insight into the limitations and potential applications of LLMs in the critical field of hate speech identification by integrating an extensive literature review with an empirical investigation.
https://arxiv.org/abs/2403.08035