Subhashish Panigrahi, Opino Gomango and Kimmi Pal
ARK https://n2t.net/ark:86534/atp-a-000002
This framework is for anyone archiving oral knowledge: how to plan a recording, how to work with a community to discuss and agree on what happens to it, and how to document that process itself. It builds on OpenSpeaks Archives' earlier work, a 2024–2025 pilot, a 2025–2026 implementation phase, and case study analysis.1
This framework speaks to three kinds of documenters: a community member recording their own language and culture, a neighbouring Indigenous documenter working with a language close to their own, and an external non-Indigenous documenter, such as a researcher, volunteer, NGO worker, or Wikimedian. The steps below are broadly the same for all three, but responsibility and power are not distributed the same way, so each role carries its own extra considerations, covered further down.
Two sets of principles run through this framework. FAIR, Findable, Accessible, Interoperable, Reusable,2 helps organise recordings and metadata, so they stay useful now and later. CARE, Collective Benefit, Authority to Control, Responsibility, Ethics,3 makes sure the recording process and its output respect Indigenous rights, community priorities, and lived experience.
FAIR is about making digital files and their descriptions easy to find, access, combine, and reuse. Findable means people can discover a recording through a simple search, because it has a clear title, language name, and keywords. Accessible means that once someone finds a file, they know how to request or access it, whether access is open or restricted. Interoperable means files and their metadata use formats and vocabularies that different tools and platforms, including Wikimedia projects, can actually work with. Reusable means a file carries a clear licence and enough description of its own origin that someone else can reuse it without guessing.
CARE is about people and relationships, centring Indigenous and other marginalised communities alongside the data itself. Collective Benefit means documentation should help the whole community first, and the project, researcher, or platform behind it second. Authority to Control means the community decides what is recorded, how it is described, how it is shared, and who can reuse it. Responsibility means a documenter acts with care, follows local protocol, and supports the community's own capacity to do this work. Ethics means documentation and its reuse respect dignity, safety, and local values; not everything that can be published should be published.
These five steps apply to whatever your role:
A community member is documenting from within. You likely already know local protocol, but it is still worth asking elders and others what is appropriate for this particular recording. Plan how other community members will be involved in reviewing, subtitling, and using the material, so the benefit is shared, not kept by one person.
A neighbouring Indigenous documenter may share a similar history of oppression and erasure with the community they are working with, but is still a guest there. Make sure local leaders and elders invite you and guide the process, and pay attention to how protocol differs: do not assume your own community's practices automatically transfer.
An external non-Indigenous documenter carries more responsibility because of the power difference between you and the community. Do not design the project alone: co-design it with community members, share decision-making with them, and be ready to step back if the community does not want a particular kind of documentation or publication.
Explain again before you record. Repeat the key points, the topic, the purpose, where the recording may be shared, and the interviewee's right to stop at any time.
Respect pauses and refusals. If someone does not want to answer or seems uncomfortable, change the question or stop altogether. Not answering is always an acceptable answer.
Co-create the questions. Where possible, ask the interviewee or community members what they want to be asked, and let them suggest topics, corrections, or a different angle.
Include context. Record a short segment on who is speaking, where, when, and why the topic matters. This helps whoever interprets or reuses the recording later.
Use clear, easy-to-understand file names. For example: LanguageCode_Village_Topic_SpeakerInitials_YYYYMMDD.extension.
Record good audio and video. Use a stable device, minimise background noise where you can, and make sure everyone who wants to be clearly seen or heard actually is, unless they have asked not to be.
Capture basic metadata while recording. Note the language or languages, the location (at whatever level of detail the community is comfortable sharing), the date, and the key themes, whether in a notebook, a form, or an app.
Think about accessibility from the start. Where you can, record in a way that leaves room to add subtitles, transcripts, or translations later; shorter segments are usually easier to subtitle and reuse.
Community review (CARE):
Metadata and licences (FAIR):
Write rich but respectful descriptions, including language, a content summary, keywords, and any cultural context the community has approved. Use licences the community has actually agreed to: for Wikimedia, that often means Creative Commons Attribution-ShareAlike (CC BY-SA), but only once the community understands and accepts what that implies; a community may allow some recordings under CC BY-SA and not others, and that choice should be respected file by file. Connect recordings to other data where it makes sense, such as related recordings, Wikipedia articles, or other references, since this makes a collection more interoperable and reusable. Keep metadata even when a file itself stays restricted: a minimal description, with the community's approval, lets others know the resource exists and how to request access, without exposing the file itself.
As a community member, decide together which recordings can go to Wikimedia Commons or another open repository, and use subtitles, transcripts, and translations to make the content usable across different Wikimedia projects: Wikipedia articles, Wikidata items, or Wikisource, where a transcript has been peer-reviewed. Treat Wikimedia as only one layer of your archive; keep or build a community-controlled copy first.
A neighbouring Indigenous documenter should check again that publishing on a global platform still matches what the recorded community actually wants. Where you are acting as a bridge to Wikimedia, keep local leaders involved in decisions about article content, image choice, and framing, and share back which pages or entries were created or improved because of the recordings.
If you are an external non-Indigenous documenter, make sure local learning resources, language classes, or community archives are part of the project design from the very beginning. Do not rush to put everything online: prioritise what the community explicitly wants on Wikimedia, and what they feel is framed correctly. In academic or public writing, cite the community and interviewees properly, and share drafts with them for feedback before publishing, wherever that is possible.
If you are a community member:
If you are a neighbouring Indigenous documenter:
If you are an external non-Indigenous documenter:
This compensation model rests on the FAIR and CARE principles above, together with three further ideas. The Equity-Centered Participatory Compensation Model4 centres marginalised voices and fair pay. Reparative justice means acknowledging historical oppression and the systemic barriers it left behind. Intersectionality means recognising that a person can face more than one kind of marginalisation at once, and that these compound rather than simply add up.
The table below gives example rates only. Fix the actual amount to your own local labour system and cost of living; do not copy these numbers directly.
| Role | Base rate (USD/hour) | Reason |
|---|---|---|
| Native speaker interviewee | $25–35 | Knowledge holder; primary contributor |
| Subtitle creator or translator | $20–30 | Skilled linguistic work; bilingual or multilingual |
| Community reviewer | $20–25 | Quality assurance; cultural verification |
| Language coordinator | $25–35 | Project management; community liaison |
| Basic technical support | $18–25 | Basic technical help |
A person can carry more than one layer of marginalisation at once. Treat these as compounding, not as separate line items to add up by hand: someone who is both Dalit and a woman, for instance, faces both barriers together, and a fair rate should reflect that combination, not just the larger of the two alone.
Three layers matter most. Historical marginalisation covers caste, tribe, race, and religion, for example, a Dalit, Adivasi, or OBC contributor, or someone from a persecuted religious minority such as Dalit-Christians in India, Muslims in India, Myanmar, or the United States, Christians in Pakistan, or Tamil Hindus in Sri Lanka. Gender-based marginalisation covers women and femme-identifying contributors, trans and non-binary contributors, and widows working in patriarchal contexts. Disability and health covers physical and sensory disability, neurodivergence, and chronic illness. Each layer typically raises the base rate by between 15% and 30%, weighted by how severe and long-standing the particular barrier is.
Use the base rate table as your starting point, then raise it deliberately for each layer a person carries, using the categories above as a guide to which factors matter and roughly how much weight each deserves. Work out the exact figure with a dedicated calculation tool built on this same approach, rather than by hand from a printed table. What this appendix gives you is the principle: pay more for compounded marginalisation, never less, and never flatten someone's circumstances into a single rate that ignores what they are carrying.
A community reviewing its own recordings of an Indigenous language can work much like peer review does in scientific publishing: a protocol for checking, correcting, and approving what has been recorded. Documenting this process is worth doing for its own sake. Record which community members reviewed the material, what concerns or corrections they raised, how any disagreements were discussed and resolved, and what the group finally agreed on. Kept alongside the recording, this becomes evidence that the material was not only made but also checked and approved, in the same way editorial review notes accompany a conventional publication.
The assessment itself has to follow local protocol and local ways of knowing: who is allowed to review particular material, what must stay private, and how authority and expertise are actually recognised in that community.
The goal here is to prepare oral knowledge recordings so they can be cited. Where resources allow, give each recording a transcript and subtitles in at least one widely used language, ideally a neighbouring official language that native speakers are already comfortable with; this lets people outside the community understand the content too. See Appendix A for the practical mechanics of captioning and subtitling.
Transcription quality needs review as well: community members can check a transcript and its subtitles the same way they review the original recording (see B.4 above), adding another layer of verification.
For citation, a time-coded transcript can work much like page numbers do in a printed text: each statement can be linked to a specific moment in the recording. This gives editors and readers a concrete way to verify what was said and where, and pushes back against the idea that oral sources are somehow less precise than written ones.
Use this framework throughout your documentation work, from the first conversation with a community through to sharing recordings publicly.
Available in: English