SARCASM DETECTION IN TEXT USING DEEP LEARNING NETWORKS
Mushafaq Yousuf Ashai, Kamal Kumar · Journal of Emerging Technologies and Innovative Research · 2021
Textual sentiment or opinion analysis systems mine textual data to identify personal feelings or views about a particular item or event. However, if sarcastic features of conversation are not taken into account, then these systems may be biased. As a result, sarcasm detection in textual communication is important for these systems' performance. Several research have used numerous methods to identify sarcasm in text, but all lack a critical component of any textual form of communication: context and semantics. The context and semantics are captured using BERT Model. Then, the classifiers are subsequently trained using these rich context and semantic embeddings. Using two datasets, we compared our system to state-of-the-art systems and found that BERT had higher F1-score, recall, and precision. As a result, we conclude that incorporating contextual and semantic data into sarcastic classifiers increases their overall performance.