Detection of Hate Speech in Hinglish Language

Abstract

As mobile phones and Internet become more and more popular, the number of social media users in India continues to go up. Majority of Indian social media users use Hinglish as their medium of communication. The Hinglish language is a mixture of Hindi words (typed in English) and English words. However, with increasing numbers, there is also an increase in the amount of hate-filled messages, posts, and comments put up on social media platforms. Hate speech is usually done to target an individual or group of individuals on the basis of caste, community, ethnicity, religion, gender, or any other discriminating factor. It can have negative impacts on the individuals facing it and consequently on the society as well. As the amount in which such kind of content is generated is huge, it becomes necessary to automatically detect hate speech so that preventive measures can be taken to control it. Although there has been quite a lot of research on hate speech detection in English texts, not much work can be found on hate speech detection in Hinglish language. This paper presents an approach of detecting hate speech in Hinglish texts using long short-term memory (LSTM), which works on word embeddings generated by gensim’s word2vec model.