# 🧠 SentimentClassifier-BERT-RegulatoryCompliance

A **BERT-based** sentiment analysis model fine-tuned on regulatory feedback and compliance-related text. This model classifies input text into **Positive**, **Neutral**, or **Negative**, making it well-suited for analyzing complaints, formal feedback, and regulatory communication.

---

## ✨ Model Highlights

- 📌 Based on [`bert-base-uncased`](https://huggingface.co/bert-base-uncased)
- 🔍 Fine-tuned on a custom dataset of labeled regulatory feedback
- ⚡ Supports prediction of **3 classes**: Positive, Neutral, Negative
- 🧠 Built using **Hugging Face Transformers** and **PyTorch**

---

## 🧠 Intended Uses

- ✅ Regulatory and compliance feedback classification
- ✅ Complaint monitoring and triaging
- ✅ Customer sentiment analysis for compliance departments

---

## 🚫 Limitations

- ❌ Not optimized for multi-language input (English only)
- 📏 Input longer than 128 tokens will be truncated
- 🤔 Model may misinterpret informal or slang language
- ⚠️ Not intended to replace expert human judgment in legal matters

---

## 🏋️‍♂️ Training Details

| Attribute          | Value                              |
|-------------------|------------------------------------|
| Base Model         | `bert-base-uncased`               |
| Dataset            | Custom `.txt` file with feedbacks |
| Labels             | Negative (0), Neutral (1), Positive (2) |
| Max Token Length   | 128                                |
| Epochs             | 3                                  |
| Batch Size         | 16                                 |
| Optimizer          | AdamW                              |
| Loss Function      | CrossEntropyLoss                   |
| Framework          | PyTorch + Transformers             |
| Hardware           | CUDA-enabled GPU                   |

---

## 📊 Evaluation Metrics

| Metric    | Score |
|-----------|-------|
| Accuracy  | 0.84  |
| Precision | 0.85  |
| Recall    | 0.84  |
| F1 Score  | 0.85  |

---

## 🔎 Label Mapping

| Label ID | Sentiment |
|----------|-----------|
| 0        | Negative  |
| 1        | Neutral   |
| 2        | Positive  |

---

## 🚀 Usage

```python
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
import torch.nn.functional as F

model_name = "your-username/sentiment-bert-regulatory-compliance"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)
model.eval()

def predict(text):
    inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True, max_length=128)
    with torch.no_grad():
        outputs = model(**inputs)
        probs = F.softmax(outputs.logits, dim=1)
        pred = torch.argmax(probs, dim=1).item()
        label_map = {0: "Negative", 1: "Neutral", 2: "Positive"}
        return f"Sentiment: {label_map[pred]} (Confidence: {probs[0][pred]:.2f})"

# Example
print(predict("The issue was resolved promptly and professionally."))
```

## 📁 Repository Structure
```
bash
Copy
Edit
.
├── model/               # Fine-tuned model files (pytorch_model.bin, config.json)
├── tokenizer/           # Tokenizer config and vocab
├── training_script.py   # Training code
├── feedbacks.txt        # Source dataset
├── README.md            # Model card
```
## 🤝 Contributing

Contributions are welcome! Feel free to open an issue or pull request to improve the model or its documentation.