emtract distilbert base uncased emotion

Introduction

The emtract-distilbert-base-uncased-emotion model is a fine-tuned version of DistilBERT designed to detect emotions in financial social media content. It is particularly effective in analyzing messages from platforms like StockTwits, categorizing them into seven emotion types: neutral, happy, sad, anger, disgust, surprise, and fear.

Architecture

The model is based on the DistilBERT architecture, which is a smaller, faster version of BERT. It has been adapted for financial emotion analysis, taking into account specific nuances of financial discussions on social media platforms.

Training

The training process involved a blend of datasets, starting with the Unify Emotion Datasets and supplemented by a custom dataset of 10,000 messages from StockTwits. The model's hyperparameters included a sequence length of 64, a learning rate of 2e-5, and a batch size of 128, over 8 epochs. Evaluation metrics used were accuracy, precision, recall, and F1-score.

Guide: Running Locally

1. Clone the repository containing the model and its dependencies.
2. Install the necessary libraries using pip, such as PyTorch and the Transformers library from Hugging Face.
3. Load the model using the provided scripts or Jupyter Notebook (Inference.ipynb) for inference tasks.
4. Run the model on local data or examples to see emotion classification results.

For large-scale tasks or faster performance, consider using cloud GPU providers such as AWS, Google Cloud, or Azure to leverage their computational power.

License

This project is licensed under the MIT License, allowing for wide use and modification with proper attribution.