--- pretty_name: Nepal Oral Speech (Research) language: - ne - en - multilingual license: other task_categories: - automatic-speech-recognition - audio-classification - text-to-speech tags: - nepal - oral-languages - research - gated - language-documentation - speech-corpus size_categories: - n<1K extra_gated_prompt: >- This dataset contains research-licensed oral language recordings from Nepal-related community projects. Access is limited to research and revitalization uses consistent with partner MoUs. No redistribution, no commercial resale, no training of public commercial models without written approval. You agree not to attempt re-identification of speakers and to delete data if access is revoked. extra_gated_fields: Affiliation: text Research purpose: text I agree to the research terms: checkbox --- # Nepal Oral Speech (Research) **Gated research speech corpus** for approved researchers and partner techs ([AkAiNp](https://huggingface.co/AkAiNp)). - Clips with backbone license **`research`** only - Includes self and guardian modes when research license was chosen - Rich metadata for dialect / origin / consent versions - **Not** a substitute for the private API vault (`private` clips never land here by default) Public users should use [`nepal-oral-packs-open`](https://huggingface.co/datasets/AkAiNp/nepal-oral-packs-open) and [`nepal-oral-demo`](https://huggingface.co/datasets/AkAiNp/nepal-oral-demo) instead. ## Access 1. Request access on the Hub with affiliation + purpose. 2. Maintainers / stewards approve per MoU. 3. On revocation: remove access; downstream users must delete local copies. This repository should be **gated** on the Hub. ## Intended uses | OK (if approved) | Not OK | |------------------|--------| | Academic ASR/TTS experiments | Publishing the raw set unauthenticated | | Partner evaluation of models | Scraping into public mirrors | | Paper results with aggregation | Commercial product training without addendum | | Internal NGO quality review | Treating research license as CC-BY | ## Data fields (target) | Field | Type | Notes | |-------|------|--------| | `clip_id` | string | | | `track_id` / `variety_id` | string | | | `mode` | string | `self`, `guardian`, `respeak`, … | | `license` | string | Always `research` in this repo | | `consent_versions` | list[string] | Template ids | | `operator_token` / `speaker_token` | string | Pseudonymous only | | `meaning_en` / `meaning_ne` | string | If linked to an item | | `audio` | Audio | Required for speech rows | | `duration_ms` / `sha256` | | | | `geo_district` | string | Approximate only; optional | | `review_status` | string | Prefer `approved` only in export | | `exported_at` | string | ISO-8601 | ## Excluded - `private` license clips - Rejected / withdrawn clips - Precise GPS, legal names, executed MoU files - Sacred / steward-blocked prompt categories ## License / terms Hub `license: other` — research agreement + MoU, not a standard open content license. ## Related - [`AkAiNp/nepal-oral-demo`](https://huggingface.co/datasets/AkAiNp/nepal-oral-demo) - [`AkAiNp/nepal-oral-packs-open`](https://huggingface.co/datasets/AkAiNp/nepal-oral-packs-open)