Sentiment Analysis and Sarcasm Detection using Deep Multi-Task Learning.

Area of Science:

• Natural Language Processing
• Artificial Intelligence
• Computational Linguistics

Background:

• Social media generates vast amounts of user-generated content valuable for sentiment analysis.
• Sentiment analysis classifies text as positive, negative, or neutral, but struggles with sarcasm.
• Sarcasm, often with negative intent masked by positive language, is a challenge for sentiment analysis.

Purpose of the Study:

• To investigate the correlation between sentiment analysis and sarcasm detection.
• To develop a unified framework that leverages this correlation to enhance sentiment analysis performance.
• To address the limitations of standalone sentiment and sarcasm classification models.

Main Methods:

• Proposed a multi-task learning framework using a deep neural network.
• Modeled the inherent correlation between sentiment analysis and sarcasm detection tasks.
• Utilized deep learning algorithms for improved text categorization.

Main Results:

• Demonstrated a significant correlation between sentiment analysis and sarcasm detection.
• The proposed multi-task learning model achieved a 94% F1-score.
• Outperformed existing standalone methods by 3%.

Conclusions:

• Jointly learning sentiment analysis and sarcasm detection improves overall performance.
• The developed deep learning framework effectively models the relationship between the two tasks.
• This approach offers a more robust solution for analyzing sentiment in text data, especially on social media.