|
|
--- |
|
|
title: 🐦 TwittBERTO |
|
|
emoji: 🚗 |
|
|
colorFrom: blue |
|
|
colorTo: indigo |
|
|
sdk: streamlit |
|
|
sdk_version: 1.42.2 |
|
|
app_file: app.py |
|
|
pinned: false |
|
|
--- |
|
|
|
|
|
|
|
|
This project demonstrates a sentiment analysis pipeline built with **DistilBERT**, a lightweight transformer model developed by Hugging Face. The model was fine-tuned on a dataset of 16,000 tweets to classify sentiment into categories such as **Positive**, **Negative**, and **Neutral**. The final model achieved an impressive **90% accuracy** on the validation set. |
|
|
|
|
|
--- |
|
|
|
|
|
## 🚀 Features |
|
|
|
|
|
* Utilizes **DistilBERT** for high-performance NLP with lower resource consumption. |
|
|
* Cleaned and preprocessed Twitter data (16K rows). |
|
|
* Fine-tuned with PyTorch and Hugging Face Transformers. |
|
|
* Achieved **90%+ accuracy** on sentiment classification. |
|
|
* Includes training, validation, and evaluation pipelines. |
|
|
|
|
|
--- |
|
|
|
|
|
## 📁 Dataset |
|
|
|
|
|
* 16,000 manually labeled tweets with three sentiment classes: |
|
|
|
|
|
* `Positive` |
|
|
* `Negative` |
|
|
* `Neutral` |
|
|
* Dataset was preprocessed to remove mentions, hashtags, links, and special characters. |
|
|
|
|
|
--- |
|
|
|
|
|
## 🧠 Model |
|
|
|
|
|
* **Base Model**: `distilbert-base-uncased` |
|
|
* **Fine-tuning**: Trained for several epochs using a cross-entropy loss function and AdamW optimizer. |
|
|
* **Tokenizer**: Hugging Face `DistilBertTokenizerFast` |
|
|
* **Training Framework**: PyTorch + Hugging Face `Trainer` API |
|
|
|
|
|
--- |
|
|
|
|
|
## 📊 Performance |
|
|
|
|
|
| Metric | Score | |
|
|
| --------- | ----- | |
|
|
| Accuracy | 90% | |
|
|
| Precision | High | |
|
|
| Recall | High | |
|
|
| F1-score | High | |
|
|
|
|
|
> Note: Actual precision, recall, and F1-score values can be added if available. |
|
|
|
|
|
--- |
|
|
|
|
|
## 📦 Dependencies |
|
|
|
|
|
```bash |
|
|
transformers==4.x.x |
|
|
torch==1.x |
|
|
scikit-learn |
|
|
pandas |
|
|
numpy |
|
|
matplotlib |
|
|
``` |
|
|
|
|
|
Install with: |
|
|
|
|
|
```bash |
|
|
pip install -r requirements.txt |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## 🛠️ How to Run |
|
|
|
|
|
1. Clone the repository: |
|
|
|
|
|
```bash |
|
|
git clone https://github.com/yourusername/twitter-sentiment-distilbert.git |
|
|
cd twitter-sentiment-distilbert |
|
|
``` |
|
|
|
|
|
2. Install dependencies: |
|
|
|
|
|
```bash |
|
|
pip install -r requirements.txt |
|
|
``` |
|
|
|
|
|
3. Train the model: |
|
|
|
|
|
```bash |
|
|
python train.py |
|
|
``` |
|
|
|
|
|
4. Evaluate the model: |
|
|
|
|
|
```bash |
|
|
python evaluate.py |
|
|
``` |
|
|
|
|
|
5. Run prediction on new tweets: |
|
|
|
|
|
```bash |
|
|
python predict.py --text "I love this app!" |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## 📈 Example Output |
|
|
|
|
|
```bash |
|
|
Input: "I love this app!" |
|
|
Predicted Sentiment: Positive |
|
|
``` |
|
|
|
|
|
--- |
|
|
|
|
|
## 📚 Future Improvements |
|
|
|
|
|
* Integrate with a live Twitter API for real-time sentiment tracking. |
|
|
* Add a web dashboard using Streamlit or Flask. |
|
|
* Extend to multilingual support using `xlm-roberta`. |
|
|
|
|
|
--- |
|
|
|
|
|
## 📄 License |
|
|
|
|
|
This project is open-source and available under the [MIT License](LICENSE). |
|
|
|
|
|
--- |
|
|
|
