Detecting Hate Speech in Hindi in Online Social Media

Abstract

Because of the rise in online hatred, the research communities of artificial intelligence, particularly natural language processing, have been developing models for identifying online hatred. Recently, code-mixing, or the usage of multiple languages in social media conversations, has made multilingual hatred a significant difficulty for automated detection. The crucial task involved in NLP is identifying inciting hatred in writings on social networking sites. This work has several relevant applications, including analysis of sentiments, cyberbullying in online world, and societal & political conflict studies. Using tweets that have been put online on Twitter, we analyze the issue of hatred detection in multilingual functionality in this paper. The tweets have the text annotations and the speech category (Normal speech or Hate speech) to which these belong. We, therefore, recommend a monitored method for detecting hatred. Additionally, the classification approach is provided, which uses certain characters level, words level, and lexicons-based features for identifying hate speech in the corpus. We obtain results of 96% accuracy in identifying posts across four classifiers. Index Terms—Hate speech, Multilingual, Code-mixing, NLP