Beyond Words: Contextual Sarcasm Detection in News Texts using Advanced Models

This research paper presents an in-depth exploration of sarcasm detection employing a synthesis of machine learning and natural language processing techniques. Drawing upon a diverse dataset amalgamated from various sources, the study assesses the efficacy of conventional models such as Random Forest and Gradient Boosting in conjunction with advanced deep learning architectures, including Long Short-Term Memory (LSTM) and Convolutional Neural Network (CNN). The research introduces an innovative ensemble model, seamlessly integrating the strengths of these approaches—CNN, LSTM, and Random Forest—with a step decay learning rate scheduler for dynamic optimization. Attention mechanisms are incorporated to enhance interpretability. The comprehensive evaluation of the proposed model reveals an outstanding accuracy of 98.56%. Furthermore, the research extends its focus to sentiment analysis, highlighting the versatility of the models. The incorporation of a learning rate schedule contributes to the adaptability and robustness of the ensemble model, showcasing its effectiveness in the evolving landscape of sentiment and sarcasm analysis within computational linguistics.