hmar-lang/hmar-language-basic-words

Hmar Dictionary (Basic Words)

A comprehensive collection of ~11,000 Hmar words and phrases, primarily sourced from Dr. John H. Pulamte's dictionary (Pherzawl Diksawnari). This dataset is optimized for NLP tasks, language preservation, and machine translation.

Repository Structure

• data/: Raw CSV source files.
• processed/: Training-ready files in JSON Lines (.jsonl) format.
  • train.jsonl: 95% of the data (~10,300 entries).
  • test.jsonl: 5% of the data (~500 entries).
• hmar_data.json: The complete dataset in a single JSON array.
• convert_to_json.py: Script used to process raw data and generate splits.

Data Structure

JSON/JSONL Format

The data is structured Hmar-first to prioritize the primary language:

{
  "hmr": "A bangkhai",
  "en": "remainder, extra, surplus",
  "pos": "v"
}

• hmr: The Hmar word or phrase.
• en: English translation or description.
• pos: Part of speech (e.g., n for noun, v for verb, adj for adjective).

Usage

To regenerate the processed files from source:

python3 convert_to_json.py

Dataset Details

• Total Entries: 10,884
• Source: Dr. John H. Pulamte’s Dictionary (Pherzawl Diksawnari).
• Format: CSV (Source), JSON/JSONL (Processed).

License

MIT. Note that while the repository is MIT, the source data from the Pherzawl Diksawnari may be subject to external copyright.