README.md

# Dataset Card for HVU_QA

**HVU_QA** is an open-source Vietnamese Question–Context–Answer (QCA) corpus and supporting tools for building FAQ-style question generation systems in low-resource languages.  
The dataset was created using a fully automated pipeline that combines **web crawling from trustworthy sources, semantic tag-based extraction, and AI-assisted filtering** to ensure high factual accuracy.

---

## Dataset Summary

- **Language:** Vietnamese  
- **Format:** SQuAD-style JSON  
- **Total samples:** 30,000 QCA triples (full corpus released)  
- **Domains covered:** Social services, labor law, administrative processes, and other public service topics  

Each entry in the dataset has the following structure:
- **Question:** Generated or extracted question  
- **Context:** Supporting text passage from which the answer is derived  
- **Answer:** Answer span within the context  

---

## Supported Tasks and Benchmarks

- **Question Generation (QG)**  
- **Question Answering (QA)**  
- **FAQ-style dialogue systems**  

A fine-tuned `VietAI/vit5-base` model trained on HVU_QA achieved:  
- **BLEU:** 90.61  
- **Semantic similarity:** 97.0% (cosine similarity ≥ 0.8)  
- **Human evaluation:**  
  - Grammaticality: 4.58 / 5  
  - Usefulness: 4.29 / 5  

---

## Languages

- **Vietnamese** (primary)  

---

## Dataset Structure

### Data Fields

Each sample contains:
- `question`: A natural language question  
- `context`: Supporting text passage  
- `answer`: The extracted answer span  

### Data Splits

| Split | Size |
|-------|------|
| Train | 30,000 |

---

## Dataset Creation

### Creation Pipeline

The dataset was built using a 4-stage automated process:
1. **Selecting relevant QA websites** from trusted sources  
2. **Automated data crawling** to collect raw QA webpages  
3. **Extraction via semantic tags** to obtain clean Q–C–A triples  
4. **AI-assisted filtering** to remove noisy or factually inconsistent samples  

---

## Usage Example

```python
from datasets import load_dataset

dataset = load_dataset("DANGDOCAO/GeneratingQuestions")
print(dataset["train"][0])
```

Example output:
```json
{
  "question": "What type of coffee is famous in Vietnam?",
  "context": "Iced milk coffee is a famous drink in Vietnam.",
  "answer": "Iced milk coffee"
}
```

---

## Training & Fine-tuning

To fine-tune a question generation model:

```bash
python fine_tune_qg.py
```

- Loads `30ktrain.json`  
- Fine-tunes `VietAI/vit5-base`  
- Saves model as `t5-viet-qg-finetuned/`  

👉 Alternatively, you can use the pre-trained model provided here:  
[Pre-trained model link](https://huggingface.co/datasets/DANGDOCAO/GeneratingQuestions/tree/main)

---

## Question Generation Example

```bash
python generate_question.py
```

**Input passage:**
```
Iced milk coffee is a famous drink in Vietnam.
```

**Generated questions:**
1. What type of coffee is famous in Vietnam?  
2. Why is iced milk coffee popular?  
3. What ingredients are included in iced milk coffee?  
4. Where does iced milk coffee originate from?  
5. How is Vietnamese iced milk coffee prepared?  

---

## Citation

If you use **HVU_QA** in your research, please cite:

```bibtex
@inproceedings{nguyen2025hvuqa,
  title={A Method to Build QA Corpora for Low-Resource Languages},
  author={Ha Nguyen-Tien and Phuc Le-Hong and Dang Do-Cao and Cuong Nguyen-Hung and Chung Mai-Van},
  booktitle={Proceedings of the International Conference on Knowledge and Systems Engineering (KSE)},
  year={2025}
}
```

---

## License

This dataset is released for **research purposes only** under the **CC BY-NC-SA 4.0 license**.