Short answer. The Puno Quechua corpus partnered with a university and a local community organisation separately, because they contribute different things and conflating them loses both. The four-phase participatory model runs planning, preparation, collection and deployment, with governance settled in phase two — before any data is collected, not after. Preparation set a CC0-1.0 licence, prepared seed content across agriculture, healthcare and technology, and localised the Mozilla Common Voice interface. Collection used voluntary skill-based contributions with community-led validation.
Key takeaways
- The Puno Quechua corpus established partnerships with both a university and a local community organisation, engaged separately because they contribute different things.
- The four-phase participatory model runs planning, preparation, collection and deployment, with governance decided in phase two before any data exists.
- Universities typically contribute methodological rigour, ethics infrastructure, linguistic expertise, trainable students and institutional continuity; community organisations contribute access, trust, variety knowledge, appropriateness judgement and validation against lived experience.
- FAIR and CARE principles, alongside PILARS, are named as the standards for long-term sustainability in national language data commons infrastructure, framed by Indigenous data sovereignty.
- Durable partnerships share two features: a named liaison responsible for the relationship, and reciprocity — material and capability flowing in both directions rather than only toward the project.
Why partner with a university and a community organisation separately?
Because a university and a community body contribute different things, and treating one as a proxy for the other collapses the distinction most commercial projects get wrong.
The Puno Quechua speech corpus offers the clearest published template for this kind of partnership. Its planning phase involved establishing partnerships with the National University of Altiplano Puno and the local community organisation Illariy Ch'aska, alongside identifying the ISO 639-3 code and assessing community needs — the item most commercial projects skip, and the one that distinguishes a partnership from a supplier relationship.
The Tonalli Corpus consortium in Mexico names the division of labour explicitly: an intercultural education university for deep understanding of indigenous languages and cultures; a technical institution for engineering and processing; INALI for linguistic accuracy across Mexico's indigenous languages; INPI for the rights and cultures of indigenous peoples; and a musical group promoting Nahuatl and other ancient languages through song. That last partner is not an obvious choice for a corpus project, and it reaches speakers, contexts and registers a university department does not.
What universities typically bring: methodological rigour, ethics review infrastructure, linguistic expertise in the specific language family, students who can be trained into the work, archival capacity and institutional continuity beyond a project cycle. What community organisations typically bring: access, trust, knowledge of which varieties matter and to whom, judgement about what is appropriate to record and release, and the ability to validate content against lived experience rather than reference works. A project with only the first produces methodologically sound data the community had no say in; a project with only the second produces culturally grounded data that may not meet technical specification. The published successes have both.
What does the four-phase participatory model look like?
The Puno Quechua team structured its work as a participatory design process — a method that builds the research plan together with the community rather than applying it to the community — across four phases: planning, preparation, collection and deployment.
Planning identifies the language precisely, including its ISO code, establishes the partnerships, and assesses community needs — the step that most often gets skipped under a commercial timeline.
Preparation sets up data governance (a CC0-1.0 licence, in this case), prepares seed sentences and questions — here spanning agriculture, healthcare and technology — and localises the collection platform, in this case Mozilla Common Voice, into Puno Quechua. Governance is decided here, before any data exists; deciding licensing after collection means renegotiating with everyone who contributed.
Collection is voluntary and skill-based, spanning reading, speaking, listening and writing, with community-led validation and privacy-preserving processing. Skill-based contribution lets people participate in the mode that suits them, widening participation beyond those comfortable being recorded, and community-led validation is both a quality decision and a governance one.
Deployment is open release, in this case on Mozilla Data Collective, with certificates of contribution and voucher incentives rather than cash alone — a reminder that in a community partnership, contribution is partly reciprocal and partly recognised rather than purely transactional. Recognition has value independent of payment, though it is not an argument against paying contributors fairly for multilingual data collection work.
Which governance frameworks now apply to this work?
Two named frameworks, FAIR and CARE, now sit alongside a broader principle of Indigenous data sovereignty, and treating language data work as good practice rather than expected practice under these frameworks is a mistake worth avoiding.
FAIR is a data standard covering findability, accessibility, interoperability and reusability. CARE is a companion standard covering collective benefit, authority to control, responsibility and ethics, created specifically because FAIR principles alone can facilitate extraction. Both are named alongside PILARS in the Language Data Commons of Australia's standards for long-term sustainability.
Indigenous data sovereignty is the framing underneath: communities seeking authority over how data about them is collected, used and shared, on the understanding that data includes culture, language, land and relationships, not just statistics. Because language is the data in these projects, they sit squarely inside the sovereignty conversation rather than adjacent to it.
One process is worth copying directly: a research team developed ethics resources and documents over several months, had them reviewed by Indigenous group leaders, and treated the result as a living document that other projects could adapt. The recommended focus areas from that review make a usable checklist — community and project context, the changing digital landscape, individual and collective knowledge protections, planned project outputs, and confidentiality and anonymity nuances.
The distinction between individual and collective knowledge protection is the one least familiar to commercial teams running standard consent processes for low-resource speech data. A song, a story or a ceremonial term may belong to a community rather than to the person who recorded it, and individual consent does not settle it — a point covered in more depth in how contributors are consented and paid.
What makes these partnerships durable?
Structure, not goodwill. Two examples — one in the United States, one in Australia — show the same two features: a named person responsible for the relationship, and material flowing in both directions.
MILPA, the Mexican Indigenous Languages Promotion and Advocacy collective, has run since 2015 as a partnership between faculty, graduate students and undergraduates at UC Santa Barbara and members of the diasporic Mexican Indigenous community in Santa Barbara and Ventura counties, most of them affiliated with the nonprofit MICOP.
CoEDL and AIATSIS in Australia established a partnership from the outset of a multi-year research centre, with a specific research associate named as liaison rather than the relationship being everyone's responsibility. The partnership worked in both directions: it improved access to collections for the research programme and supported community groups to deposit materials with the archive, providing safekeeping while making resources more accessible to Indigenous communities. A partnership where material flows only one way is an extraction arrangement with a friendly name.
What is being built to support this work now?
Two developments change the landscape for anyone entering this space: a global capacity-building programme, and national data commons infrastructure.
The New Commons Indigenous Language Data Commons Incubator, announced in 2026, is a six-month programme supporting Indigenous-led teams to build community-governed datasets — datasets structured for responsible, equitable use for public-interest purposes rather than open release or commercial licensing. It was co-designed with a 19-member global Steering Committee of Indigenous language data experts and implemented with the GovLab and Microsoft, with concept notes due in August 2026. The significance is the model, not the programme: community governance is a distinct structure being institutionalised with major backing.
National infrastructure is also emerging. The Australian Language Data Commons programme, for instance, provides governance frameworks respecting cultural protocols, support for securing at-risk language materials, and tools enabling community-led use of language data. For a commercial operation, both developments point the same way: the infrastructure and norms are being set by community and academic institutions, and the sensible position is to work within them rather than around them, a consideration worth weighing when recruiting native contributors for underrepresented languages.
How should a commercial partner approach this work honestly?
By naming a liaison, deciding governance before collection, and stating the commercial position plainly rather than letting it surface as a surprise after data has already been gathered.
Lifewood collects language data across 50+ languages, and partnership-based collection is how low-resource language work gets done — a stake worth declaring here directly. Three observations follow from the examples above.
Partnerships take longer to establish than projects take to run. MILPA is eleven years old; CoEDL and AIATSIS partnered from the outset of a multi-year centre. A commercial timeline that allocates three weeks to "establish community partnership" has misjudged the unit of time involved, and the practical fix is to build relationships ahead of demand rather than in response to a signed contract.
Name a liaison. The CoEDL model of one person responsible for the institutional relationship is worth copying exactly — a relationship that is everybody's responsibility decays quietly between projects.
Decide governance before collection, and be honest about the commercial position. A community deciding whether to work with a commercial data sovereignty-aware partner is entitled to know what happens to the data, who can license it, whether it is exclusive, and what the community retains. Those answers may be less generous than an academic open-release model, and saying so plainly beats discovering the mismatch after collection — communities tend to work more readily with a clear commercial proposition than an ambiguous one. Anyone choosing a multilingual data collection partner should expect a straight answer to each of these questions before signing anything.
What belongs on a partnership checklist?
A short, sequenced list: engage universities and community bodies separately, decide governance before collection, and build reciprocity and a named liaison into the structure rather than leaving them implicit.
- Engage universities and community bodies separately, and map what each contributes before approaching either.
- Identify the language precisely, including ISO code and target varieties, during planning.
- Assess community needs, and be prepared for the answer to change the scope.
- Decide licensing and governance during preparation, before collection.
- Localise the collection platform into the target language rather than running it in a lingua franca.
- Offer skill-based participation across reading, speaking, listening and writing.
- Make validation community-led where the content is cultural.
- Address collective as well as individual knowledge rights, since one speaker's consent does not settle community ownership of what they said.
- Build reciprocity into the structure, so material and capability flow both ways.
- Name a liaison and fund the relationship between projects, not only during them.
- Reference FAIR, CARE and Indigenous data sovereignty explicitly in the agreement, because those frameworks are now the expected vocabulary — the same vocabulary that underpins global multilingual speech data collection done responsibly.