Annushree Bablani


2024

pdf
Sentiment and sarcasm: Analyzing gender bias in sports through social media with deep learning
Sethulakshmi Praveen | Balaji Tk | Sreeja Sr | Annushree Bablani
Proceedings of the 21st International Conference on Natural Language Processing (ICON)

Gender bias continues to be a pervasive issue, especially in public discourse surrounding high-profile events like the Olympics. Social media platforms, particularly Twitter, have become a central space for discussing such biases, making it crucial to analyze these conversations to better understand public attitudes. Sentiment analysis plays a key role in this effort by determining how people feel about gender bias. However, sarcasm often complicates sentiment analysis by distorting the true sentiment of a tweet, as sarcastic expressions can mask negative or positive sentiments. To address this, the study introduces a novel framework called SENSA (SENtiment and Sarcasm Analysis), designed to detect both sentiment and sarcasm in tweets related to gender bias. The framework leverages the R2B-CNN model for robust sarcasm and sentiment classification. Using approximately 5,000 tweets related to gender bias from 2010 to August 30, 2024, SENSA applies advanced sarcasm detection to account for shifts in sentiment caused by sarcastic remarks. The R2B-CNN model demonstrates a high accuracy of 92.32% along with achieving 92.75% precision and 92.53% F1-score for sarcasm detection and a 93.67% accuracy, 92.33% precision and 92.33% F1-score for sentiment classification. SENSA provides a comprehensive understanding of gender bias discussions on social media by capturing both sentiment and sarcasm to reveal deeper insights into public perceptions.