Fine-tuning BERT for Sentiment Analysis

This project fine-tunes a BERT model for sentiment classification (positive/negative).

📌 Prerequisites

Ensure you have the following libraries installed:

pip install transformers torch scikit-learn datasets

📥 Installation & Setup

1. Clone this repository:

   git clone https://github.com/PiWarior/Sentiment_Classifier.git
   cd Sentiment_Classifier

2. Install dependencies:

   pip install -r requirements.txt

3. Download and prepare the dataset (example using IMDb):

   from datasets import load_dataset
   dataset = load_dataset("imdb")

---

🎯 Fine-tuning the Model

The BERT model was fine-tuned on a sentiment dataset for 3 epochs.

Epoch	Training Loss	Validation Loss	Accuracy	F1 Score	Precision	Recall
1	0.226500	0.270268	89.46%	0.8851	0.9726	0.8120
2	0.125300	0.261332	93.29%	0.9311	0.9568	0.9068
3	0.052100	0.280319	94.02%	0.9404	0.9384	0.9424

📊 Evaluation Results:

{
  "eval_loss": 0.2803,
  "eval_accuracy": 0.9402,
  "eval_f1": 0.9404,
  "eval_precision": 0.9384,
  "eval_recall": 0.9424,
  "eval_runtime": 793.55,
  "eval_samples_per_second": 31.50,
  "eval_steps_per_second": 0.49
}

The fine-tuned model is saved in the mon_modele_bert_finetuned/ directory, containing:

• tokenizer_config.json
• special_tokens_map.json
• vocab.txt
• added_tokens.json

---

🚀 Testing the Model

After fine-tuning, test the model using test.py:

python test.py

Example output:

Text: This movie was excellent, I loved every minute of it.
Predicted Sentiment: Positive (Confidence: 1.00)
--------------------------------------------------
Text: What a waste of time, I regret watching this movie.
Predicted Sentiment: Negative (Confidence: 1.00)
--------------------------------------------------

---

📂 Project Structure

📁 sentiment_analysis_project
 ├── mon_modele_bert_finetuned/  # Fine-tuned model
 │   ├── tokenizer_config.json
 │   ├── special_tokens_map.json
 │   ├── vocab.txt
 │   ├── added_tokens.json
 ├── test.py                     # Model testing script
 ├── train.py                     # Training script
 ├── requirements.txt             # Dependencies
 ├── README.md                    # This file

✨ Author

• PiWarior - GitHub