Nepali is commonly written in Devanagari and is used across different geographic and social contexts. A data specification should make clear which users, regions, domains, and applications the dataset is intended to serve.
Script and text conventions
Define Unicode handling, spelling normalization, punctuation, numerals, borrowed words, and tokenization requirements. If the project includes translation or classification, use examples that match the intended domain and give annotators clear guidance for ambiguous labels.
Audio and transcripts
For speech collection, clarify speaker eligibility, recording conditions, consent, audio quality, transcript format, and how to handle unclear speech or mixed-language segments. Quality checks should cover both the technical recording and whether the transcript follows the agreed conventions.
Teams sourcing Nepali language data, Nepali transcription, or Nepali annotation should include the task, language variety, modality, volume, source rights, timeline, and acceptance criteria. Langigs can discuss the scope and confirm feasibility before any delivery commitment.
