Computing and Informatics

In the field of natural language processing (NLP), detecting emotions or sentiments can be a challenging task, and sometimes emotions can be more complex than just positive or negative. However, detecting sarcasm in textual data adds another layer of complexity. Despite this, identifying the underlying sarcasm in the text has become a recent area of interest among NLP researchers. Headlines in newspapers often use sarcasm to engage readers, but readers may have difficulty recognizing it, leading to a misinterpretation of the news and spreading misinformation. As a result, there is an urgent need for technology that can automatically identify sarcasm with high accuracy. Recent studies in this domain have revealed a need for a robust and efficient model. Deep learning approaches have proven to be effective in sarcasm detection. In this work, we propose a novel two-stage model that uses a word-embedding technique to select relevant features followed by an advanced deep-learning architecture to classify sarcasm in news headlines. Our proposed method demonstrates promising results in identifying sarcasm in text with an accuracy rate of approximately 97 %. We have fine-tuned the hyper-parameters to increase the precision level, which enhances the efficacy of our model. Our work provides a significant contribution to the field of NLP by presenting a reliable and effective model for sarcasm detection. The comparison of our model with recent advancements indicates that our approach outperforms them. By using our model, readers can avoid misinterpretations and the spreading of misinformation. Therefore, our work can have a positive impact on society, and we believe that it can inspire future research in the field of sarcasm detection.