Identification of Conversational Hate-Speech in Code-Mixed Languages
Paper accepted at FIRE-2021 HASOC Track focusing on novel problem of detecting hateful and offensive comments in a conversation comprising of tweet, replies and comments where the comment or reply might not be inherently hateful but contributes to hate in context of the parent tweet. Our solution secured Rank 1 on the leaderboard with a MACRO F1 score of 0.7253 on test data.