README.md

---
language: fa
library_name: transformers
tags:
  - classification
  - legal
  - iranian-legal
  - persian
  - case-type
pipeline_tag: text-classification
---

# QomSSLab/CaseTypeClassifier-fa

**QomSSLab/CaseTypeClassifier-fa** is a Persian legal text classifier that predicts whether a court ruling (رأی) belongs to a **civil (حقوقی)** or **criminal (کیفری)** category.  
The model is designed for use in Iranian legal NLP pipelines, document organization, and downstream analysis of judicial data.

## 💡 Use Cases

- Automatic classification of Persian court rulings into civil or criminal categories.  
- Preprocessing step for legal analytics and document retrieval systems.  
- Assisting legal researchers and developers in structuring Persian legal corpora.

## 🧠 Model Details

- **Language**: Persian (Farsi)  
- **Task**: Text Classification  
- **Classes**: `civil` (حقوقی), `criminal` (کیفری)  
- **Pipeline Tag**: `text-classification`  

## 📦 Example Usage

```python
from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline

model_name = "QomSSLab/CaseTypeClassifier-fa"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

classifier = pipeline("text-classification", model=model, tokenizer=tokenizer)

text = "در این پرونده متهم به سرقت اموال عمومی محکوم شده است."
result = classifier(text)

print(result)
```

Example Output:
```python
[
{'label': 'کیفری', 'score': 0.9969141483306885}
]
```

## 📊 Evaluation

The model was trained and evaluated on a balanced dataset of Persian court rulings.  
It demonstrates high accuracy in distinguishing civil and criminal judgments.

| Metric | Value |
|:-------|:------:|
| **Training Loss** | 0.0358 |
| **Validation Loss** | 0.033996 |
| **Accuracy** | **0.9951** |
| **F1 Score** | **0.9951** |
| **Precision** | **0.9951** |
| **Recall** | **0.9951** |

✅ **Final Performance:** The model achieved **99.51% accuracy** and **0.9951 F1-score** on the validation set.

### Limitations

- Performance may degrade on highly abbreviated or informal texts.  
- Designed primarily for Iranian legal language; may not generalize to non-Iranian legal contexts.  
- Does not classify subtypes (e.g., family, property, or financial cases).