Hmar Unigrams Dataset (unigrams)

A clean, lowercased, and deduplicated vocabulary dataset containing 30,606 verified pure Hmar surface words.

Maintained as part of the Hmar Heritage Archival Project by the Hmar Heritage Foundation.

---

Dataset Summary

Category	Value
Language	Hmar (hmr)
Record Count	30,606 Verified Pure Hmar Words
Sources	Hmar Literature, Lexicons & Text Corpora
Format	1-Column JSONL ({"word": "..."})
License	MIT

---

Data Format

Each line in data/train.jsonl is a simple JSON object containing a lowercased Hmar unigram:

{"word": "abul"}
{"word": "adang"}
{"word": "hmangaina"}
{"word": "inkhawm"}
{"word": "siemthar"}

---

Extraction & Filtering Note

These unigrams were extracted directly from Hmar literature, digitized lexicons, Bibles, and song lyrics. We filtered the candidate vocabulary against a master English dictionary of 370,105 words; however, please note that some English loanwords, transliterated names, or unlisted English terms may still seep through.

---

License

MIT License. Maintained by the Hmar Heritage Foundation.