Enhancing Sarcasm Detection with Contextual Features Through Multi Layered Perceptron Model
Shyam Sunder Jannu Soloman, Nagaraju Baydeti · 2024
Sarcasm Detection is a challenging task due to its often nuanced and context-dependent nature. Sarcasm detection has emerged as a critical area in natural language processing, finding applications in sentiment analysis, social media monitoring and customer feedback analysis. The proposed model focuses on the advancing the accuracy and robustness of sarcasm detection model by integrating contextual features and employing sophisticated neural network architecture. The primary objective of the proposed work is to analyze textual data and obtain the best possible features. Features such as sarcasymbols, attenuation indicators and selective stopwords are carefully selected to enhance the model's ability to detect subtle sarcasm cues. By utilizing vector representation of textual data, extracted features and deep neural network, the proposed model identifies sarcasm from textual data. To evaluate the performance of the approaches in the proposed model, the metrics F1-Score and accuracy are used. The proposed model resulted in an impressive accuracy and F1 Score for the binary classification problem in bifurcating the text as sarcastic and non-sarcastic with 86.31 and 87.17 percentages respectively.