| --- |
| license: cc-by-4.0 |
| task_categories: |
| - text-generation |
| - fill-mask |
| - token-classification |
| language: |
| - my |
| tags: |
| - burmese |
| - wikipedia |
| - nlp |
| - dataset |
| - low-resource |
| - monolingual |
| --- |
| |
| # Burmese Only Wiki Dataset |
|
|
| This dataset is a refined, high-precision monolingual collection derived from the [DatarrX/myanmar-Wikipedia](https://huggingface.co/datasets/DatarrX/myanmar-Wikipedia) repository. It has been strictly filtered to ensure that every sentence contains exclusively Burmese characters, making it an ideal resource for language modeling, sequence-to-sequence tasks, and linguistic research where cross-lingual noise must be eliminated. |
|
|
| ## 🏛️ 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. |
|
|
| --- |
|
|
| ## 🏗️ Creation & Curation Process |
|
|
| To create this "Burmese-only" subset, we implemented a rigorous linguistic filtration pipeline: |
|
|
| 1. **Source Extraction:** Data was pulled from the processed `myanmar-Wikipedia` dataset, ensuring a consistent starting point. |
| 2. **Linguistic Purity Filtering:** We applied a strict Unicode-range filter (specifically targeting Burmese script blocks `\u1000-\u109F`) to detect and exclude sentences containing English, Chinese, Japanese, Korean, Thai, or other non-Burmese characters. |
| 3. **Symbol & Artifact Cleansing:** * Removed redundant formatting characters and markup residue (`''''''`, `||`, `##`). |
| * Eliminated incomplete punctuation artifacts and empty parenthetical groups. |
| 4. **Structural Normalization:** Enforced sentence-ending consistency by ensuring every sentence concludes with the standard Burmese full-stop (`။`). |
| 5. **Syllable Tokenization:** Every sentence was passed through the `mm-syllable` breaker, with a validation layer to remove invalid syllable combinations, ensuring the integrity of the secondary `syllable` column. |
|
|
| --- |
|
|
| ## 📊 Dataset Statistics |
| * **Total Sentences:** 353,267 |
| * **Average Sentence Length:** 106.49 characters |
| * **Average Syllable Count:** 33.84 |
| * **Structural Consistency:** 99.99% of sentences end with the standard Burmese full-stop (`။`). |
|
|
| ### Data Structure |
| | Column | Description | |
| | :--- | :--- | |
| | `text` | The clean, purely Burmese sentence. | |
| | `syllable` | The same sentence segmented into syllables with spaces. | |
|
|
| **Sample Entry:** |
| ```json |
| { |
| "text": "ဝီကီဟု ခေါ်သော ဝက်ဘ်ဆိုက် ပုံစံတစ်မျိုးကို အသုံးပြု၍ ပူးပေါင်းရေးသားခြင်းကို အဆင်ပြေစေရန် စီမံထားခြင်း ဖြစ်ပါသည်။", |
| "syllable": "ဝီ ကီ ဟု ခေါ် သော ဝက်ဘ် ဆိုက် ပုံ စံ တစ် မျိုး ကို အ သုံး ပြု ၍ ပူး ပေါင်း ရေး သား ခြင်း ကို အ ဆင် ပြေ စေ ရန် စီ မံ ထား ခြင်း ဖြစ် ပါ သည် ။", |
| "text_len": 113, |
| "syllable_count": 35 |
| } |
| ``` |
|
|
| --- |
|
|
| ## 👨💻 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. |
|
|
| If you utilize this dataset in your work, please cite as: |
|
|
| ```bibtex |
| @misc{datarrx_burmese_only_2026, |
| author = {Khant Sint Heinn}, |
| title = {Burmese Only Wiki Dataset}, |
| 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)} |
| } |
| |
| ``` |
