Data Sources
This subset isolates Latin-script Nepali from the colloquial corpus.
## Provenance Notes
- The release is built from the local preprocessing pipeline in this repository.
- Source-specific filtering and deduplication are applied before export.
- The resulting parquet file is intended for research, evaluation, and dataset inspection.