Hilarious or Hidden? Detecting Sarcasm in Hinglish Tweets using BERT-GRU

Sarcasm is the manner of usage of words that contradict the actual context of a sentence. Sarcasm detection in multilingual contexts poses unique challenges due to the complex interplay of languages and cultural nuances. Inside this research paper, we propose a novel approach for sarcasm detection in Hinglish, a code-mixed language blend of Hindi and English. Our approach combines the power of BERT (Bidirectional Encoder Representations from Transformers) and GRU (Gated Recurrent Unit) neural networks to capture the contextual and sequential information necessary for accurate sarcasm detection. We evaluate our approach on a code-mixed dataset, comparing it with several baseline models and achieving remarkable results. The proposed BERT-GRU model achieves an impressive accuracy, precision, recall, and F1-score of 0.96, bespeaking its effectiveness in sarcasm detection in Hinglish code-mixed text.