Twitter Emotion Recognition

Sentiment Analysis on Twitter data

Quick Summary

  • Performed sentiment analysis on 20K emotion dataset from Hugging Face to identify user emotions for improving personalized content and mental health support services in Python
  • Employed oversampling techniques to correct class imbalance and fine-tuned the pre-trained BERT model via data parallelism on a multi-GPU cluster with PyTorch, achieving an average F1 score of 0.91

Project Report

For more detailed project information, please visit the GitHub repo: https://github.com/Marklong7/NLP_emotion_classification