freococo/myanmar_wikitionary_words

📚 Myanmar Wiktionary Words

A high-quality dataset of over 26,000 Myanmar (Burmese) words scraped from my.wiktionary.org, enriched with subword-level segmentation. This dataset is intended for use in lexical modeling, NLP tokenization tasks, and morphological analysis of the Myanmar language.

---

✨ Dataset Highlights

- ✅ **26K+ Myanmar lexical entries**, sourced from online Wiktionary
- 🧱 **Subword segmentation** included for each word using intelligent Python rules
- 🧼 **Manual curation** to correct common Myanmar typographic issues (e.g., `န့` → `န့်`)
- 🔡 **Preserved** numeric and punctuation entries (e.g., `0–9`, `။`, `၊`) due to their semantic and grammatical importance
- 🧹 **Filtered out** isolated diacritics such as `ါ`, `ာ`, `ိ`, `ု`, etc., which carry no meaning on their own when detached

---

📦 File Format

• File: my_wikitionary.csv
• Encoding: UTF-8
• Fields:
  • my_word: the full Myanmar lexical item
  • my_subword: a space-separated list of quoted subwords (e.g., "xxx" "yyy")

📋 Example Rows

| my_word     | my_subword              |
|------------|--------------------------|
| ဧတဒဂ်      | "ဧ" "တ" "ဒဂ်"           |
| ကကဏ္ဋက    | "က" "ကဏ္ဋ" "က"         |
| ၅          | "၅"                      |
| ။         | "။" (Myanmar full stop)   |

---

🧠 Construction Process

1. Data Source: Scraped systematically from the my.wiktionary.org interface.
2. Unicode Normalization: Normalized all entries to NFC form for consistency.
3. Subword Tokenization: Employed rule-based segmentation to split compounds, prefixes, and suffixes.
4. Manual Corrections:
   • Fixed corrupted forms like န့, င့်, မ့်, ည့်
   • Preserved meaningful characters like ။ (Myanmar full stop), ၊ (comma), and numerals (၀–၉)
   • Removed meaningless standalone signs like ါ, ာ, ိ, ု which do not function as words

---

📊 Applications

- 📖 Myanmar language lexicon building
- ✂️ Subword tokenization model training
- 🧪 Morphological parsing and grammar analysis
- 🔤 Dictionary and spell-checker development
- 🤖 Preprocessing pipelines for Burmese NLP tasks

---

📁 File Statistics

- Entries: ~26,000+
- Fields: `my_word`, `my_subword`
- Format: Clean UTF-8 CSV

---

🙏 Acknowledgments

This project would not be possible without the collective care and wisdom of those who came before.

✨ While I applied only limited effort to clean and organize this dataset, all credit belongs to the tireless souls of Wiktionary—especially the great and compassionate contributors who gave countless unpaid hours to preserve and document their language.

Wiktionary, like its sister project Wikipedia, was launched by the Wikimedia Foundation in 2002—not by any single founder, but as a bold invitation to all. Proposed by Daniel Alston and technically supported early on by Brion Vibber, it has grown entirely thanks to a global community of volunteers.

These anonymous and often unsung contributors—linguists, students, monks, migrants, hobbyists—quietly type and revise, not for reward, but out of love for their language and a hope that it may be known, spoken, and lived.

Their invisible labor is a gift. A monument built stone by stone, not by one sculptor, but by thousands of hands.

• 🧭 Source: my.wiktionary.org
• 💻 Python processing and validation by Myanmar language researchers
• ❤️ Eternal gratitude to Wiktionary’s volunteers who help languages live on, one entry at a time

> _“May the value of their invisible work echo far beyond the lines of code we wrap it in.”_ 🕊️

---

🔓 License

Distributed under the **MIT License** – free to use, modify, and distribute with attribution.

---

💬 Citation

``

@dataset{freococo2025myanmarwikitionary,
  title     = {Myanmar Wikitionary Words},
  author    = {freococo},
  year      = {2025},
  url       = {https://huggingface.co/datasets/freococo/myanmar_wikitionary_words},
  note      = {Version 1.0, Hugging Face Datasets}
}