README.md

---
license: mit
language:
  - ny
  - en
pretty_name: Chichewa Text-to-SQL
size_categories:
  - n<1K
task_categories:
  - table-question-answering
tags:
  - text-to-sql
  - chichewa
  - low-resource-language
  - semantic-parsing
  - nlp
  - malawi
  - sql
  - qlora
  - few-shot
---

# Chichewa Text-to-SQL

The **first structured Text-to-SQL benchmark for Chichewa**, a low-resource Bantu language spoken by over 12 million people in Malawi and neighboring regions.
The dataset contains **400 manually curated natural language–SQL pairs** in both **Chichewa (Nyanja)** and **English**, grounded in a unified relational SQLite database covering five real-world domains from Malawi.

---

## Dataset Summary

This benchmark was constructed to investigate the adaptation of Large Language Models (LLMs) for Text-to-SQL generation in Chichewa. It supports systematic evaluation of zero-shot, few-shot (random and retrieval-augmented), and parameter-efficient fine-tuning (QLoRA) approaches for low-resource semantic parsing.

Key findings from the accompanying research:
- **English** zero-shot execution accuracy: 20% → 50% (random few-shot) → 70% (RAG few-shot) → **76.7% (QLoRA)**
- **Chichewa** zero-shot execution accuracy: 0% across all models → 41.7% (RAG few-shot) → **41.7% (QLoRA)**

---

## Database Schema

The SQLite database (`database/chichewa_text2sql.db`) contains five tables:

| Table | Description |
|---|---|
| `production` | Agricultural crop yield by district and season |
| `population` | Census data with geographic and demographic breakdowns |
| `mse_daily` | Malawi Stock Exchange daily trading data |
| `commodity_prices` | Commodity price data across markets |
| `food_insecurity` | Food insecurity indicators by region |

---

## Dataset Structure

### Files

| File | Description |
|---|---|
| `data/all.json` | Full dataset (400 examples) |
| `data/train.json` | Training split |
| `data/dev.json` | Development/validation split |
| `data/test.json` | Test split |
| `data/human_translations.csv` | Human-verified translations |
| `data/split_verification.json` | Split integrity verification |
| `database/chichewa_text2sql.db` | SQLite database |
| `database_tables_csv/` | Raw CSV files for each table |

### Data Fields

Each example contains:

```json
{
  "id": 1,
  "question_en": "Which district produced the most Maize",
  "question_ny": "Ndi boma liti komwe anakolola chimanga chambiri?",
  "sql_statement": "SELECT district, MAX(yield) AS max_yield FROM production WHERE crop = 'Maize';",
  "sql_result": "[('Lilongwe', 444440.0)]",
  "difficulty_level": "easy",
  "table": "production"
}
```

### Difficulty Levels

- `easy` — single table, simple SELECT / WHERE / ORDER BY
- `medium` — aggregations, GROUP BY, LIMIT
- `hard` — multi-condition queries, subqueries, JOINs

---

## Splits

| Split | Size |
|---|---|
| Train | ~280 |
| Dev | ~60 |
| Test | ~60 |
| **Total** | **400** |

---

## Usage

```python
import json

with open("data/train.json") as f:
    train = json.load(f)

print(train[0]["question_ny"])    # Chichewa question
print(train[0]["question_en"])    # English question
print(train[0]["sql_statement"])  # Ground-truth SQL
```

---

## Citation

If you use this dataset, please cite:

```bibtex
@dataset{eze2026chichewa,
  author    = {Eze, John Emeka and Matekenya, Dunstan and Matthewe, Evance},
  title     = {Chichewa Text-to-SQL: A Low-Resource Benchmark for Semantic Parsing in Chichewa},
  year      = {2026},
  publisher = {HuggingFace},
  url       = {https://huggingface.co/datasets/johneze/chichewa-text2sql}
}
```

---

## License

[MIT](https://opensource.org/licenses/MIT)