README.md

---
license: cc-by-4.0
task_categories:
- text-generation
- fill-mask
- token-classification
language:
- my
tags:
- burmese
- wikipedia
- nlp
- dataset
- low-resource
size_categories:
- 100K<n<1M
---

# Myanmar Wikipedia Dataset (20260501)

This dataset contains a cleaned, processed, and high-quality collection of Burmese Wikipedia articles, curated to serve as a robust foundation for Natural Language Processing (NLP) and Artificial Intelligence development in the Burmese language.

## 🏛️ About DatarrX
**DatarrX (Burmese: ဒေတာအက်စ်)** is a non-profit open-source foundation dedicated to building a robust digital foundation for the Burmese language in the AI era. We believe that high-quality data is the core of AI innovation. Our mission is to empower the Burmese AI ecosystem by building high-quality open-source datasets, essential AI tools, and localized technical resources.

---

## 🏗️ Dataset Creation Pipeline

This dataset was developed through a rigorous multi-stage cleaning and preparation process to ensure maximum utility for researchers and developers:

1.  **Data Extraction:** The source material was derived from the official [Burmese Wikipedia Database Dump](https://dumps.wikimedia.org/mywiki/) (2026-05-01).
2.  **Markup Sanitization:** We systematically removed MediaWiki markup, HTML tags, template artifacts (`{{...}}`), and internal links (`[[...]]`).
3.  **Filtering & Noise Reduction:** * Applied strict Zawgyi-to-Unicode verification to ensure data consistency.
    * Removed tabular metadata, non-informative CSS/module snippets, and malformed rows.
    * Filtered out rows containing non-Burmese characters (where applicable) and low-information symbols.
4.  **Sentence Segmentation & Normalization:** * Articles were split into individual sentences. 
    * We enforced a standard Burmese full-stop (`။`) at the end of every sentence to ensure structural consistency.
5.  **Syllable Tokenization:** Using `mm-syllable`, every sentence was segmented into its constituent Burmese syllables, providing a secondary column (`syllable`) for syllable-level model training.

---

## 📊 Dataset Statistics
* **Total Sentences:** 546,045
* **Average Sentence Length:** 122.98 characters
* **Average Syllable Count:** 39.11
* **Structural Consistency:** 99.90% of sentences end with the standard Burmese full-stop (`။`).

### Data Structure
| Column | Description |
| :--- | :--- |
| `text` | The clean, normalized Burmese sentence. |
| `syllable` | The same sentence segmented into syllables with spaces. |

**Sample Entry:**
```json
{
  "text": "အကြောင်းအရာပေါင်း ခုကို ဖြင့် ဖတ်ရှုနိုင်ပါသည်။",
  "syllable": "အ ကြောင်း အ ရာ ပေါင်း ခု ကို ဖြင့် ဖတ် ရှု နိုင် ပါ သည် ။",
  "text_len": 47,
  "syllable_count": 14
}
```

---

## 👨‍💻 Creator & Contributors

This dataset is curated and maintained by **Khant Sint Heinn (Burmese: ခန့်ဆင့်ဟိဏ်း, Kalix Louis)**.

Khant Sint Heinn is a Machine Learning Engineer specializing in NLP and open-source AI development. As the Lead Developer at **DatarrX**, he focuses on building scalable data pipelines and practical infrastructure to improve support for low-resource languages. His work aims to transform limited language resources into practical opportunities through clean data and community-driven innovation.

---

## ⚖️ License & Citation

This work is released under the **Creative Commons Attribution 4.0 International (CC-BY-4.0)** license, ensuring openness and collaboration for the global AI research community.

If you find this dataset useful for your research or applications, please cite it as:

```bibtex
@misc{datarrx_my_wiki_2026,
  author       = {Khant Sint Heinn},
  title        = {yanmar Wikipedia Dataset (20260501)},
  year         = {2026},
  publisher    = {Hugging Face},
  organization = {DatarrX},
  howpublished = {https://huggingface.co/datasets/DatarrX/myanmar-Wikipedia},
  note         = {Published under DatarrX. Open-source community asset released under Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0)}
}

```