Sentimental Analysis of Twitter using Semi-Supervised Approaches

📌 Key Contributions

• Developed a sentiment analysis model to classify tweets into positive, negative, and neutral categories
• Addressed limited and noisy labeled data by applying semi-supervised learning techniques, including self-training, co-training, and lightweight ensemble methods
• Initiated training with a small hand-labeled dataset, and iteratively expanded it by automatically labeling high-confidence tweets
• Refined the classifier through multiple training iterations, improving accuracy and robustness on real-world social media data

💻 Code

The source code of this work are publicly available: [Code]