Hybrid Deep Learning Model for Sarcasm Detection in Indian Indigenous Language Using Word-Emoji Embeddings

Automated sarcasm detection is deemed as a complex natural language processing task and extending it to morphologically-rich free-order dominant indigenous Indian Hindi another challenge in itself. The scarcity of resources tools such annotated corpora, lexicons, dependency parser, Part-of-Speech tagger, benchmark datasets engorge the linguistic challenges low-resource languages like Hindi. Furthermore, context incongruity imperative detect sarcasm, various linguistic, aural visual cues can be used predict target utterance sarcastic. While pre-trained word embeddings capture meanings, semantic relationships different types contexts form representations, emojis also render useful contextual information, analogous human facial expressions, for gauging sarcasm. Thus, goal this research demonstrate use hybrid deep learning model trained using two embeddings, namely emoji validated on tweets dataset, Sarc-H, manually with sarcastic non-sarcastic labels. preliminary results clearly depict importance detection, our attaining an accuracy 97.35% F-score 0.9708. validates that automated feature engineering facilitates efficient repeatable predictive detecting indigenous, languages.