A chrome monolith towering over many smaller forms, one glowing faintly purple.

LINGUISTICS EXCELLENCE CENTRES

7,000+ languages. Each one a way of thinking.

Help build the Universal Metalinguistic Framework, the structured description of how languages differ that makes AI correction possible.

HOW A PROFILE IS BUILT

One language. Many kinds of expert.

Evidentiality, honorifics, kinship carried inside the verb: when AI drops these it does not simplify a language, it erases a worldview. Describing them takes more than one kind of knowledge.

The language profile

Linguists

describe the structure formally.

Native speakers

confirm what is actually said.

Translators

show where machine output breaks.

Anthropologists

carry the cultural and pragmatic context.

Universities

validate, publish and hold the record.

No single expert can describe a language completely. The framework is built to hold all of them, weight their contributions differently, and consolidate them into one profile.

That profile then governs how AI behaves in the language.

This is not archival work. What you describe is what the correction layer enforces, in production, in every system that uses it.

A crowd of small dark forms, several glowing faintly purple.

Your language is probably already waiting.

2,400+ languages carry baseline structural profiles in the Mātr system today. Most contributors arrive to find their language already described at the baseline, ready to be deepened by someone who actually speaks it.

Find your language

CURRENTLY BEING DEEPENED

Languages being deepened right now.

These are languages where contributors are actively adding features, values, and examples on top of a baseline profile. The wall reflects work in progress, not finished descriptions.

  • বাংলাBengali
  • नेपालीNepali
  • ગુજરાતીGujarati
  • 한국어Korean
  • བོད་སྐདTibetan
  • SoomaaliSomali
  • WolofWolof
  • CymraegWelsh
  • GaeilgeIrish
  • EuskaraBasque
  • ᏣᎳᎩCherokee
  • Diné bizaadNavajo
  • Na Vosa VakavitiFijian
  • 𒀀𒀝𒂍Akkadian
  • 𒅴𒂠Sumerian
  • کٲشُرKashmiri
  • тыва дылTuvinian
  • Dorerin NaoeroNauru
  • Gagana SamoaSamoan

WHY THIS COMPOUNDS

This is not data collection. It is framework building.

Every feature you describe enters the Universal Metalinguistic Framework, the same structure that drives correction in production. Your description of how your language marks evidence, or ranks honorifics, or orders a relative clause, becomes part of how the system reasons about every language that shares that property.

That is why this compounds. A corpus improves one language. A framework entry improves the treatment of a whole structural pattern, wherever it occurs.

How contribution works.

01

Apply and verify

Tell us your languages and your background. We review every application personally.

02

Adopt features

Choose the parts of your language you know best. Nobody is asked to describe everything.

03

Contribute values

Provide the feature value, an example in native script, a gloss, and your evidence. Confidence is recorded alongside every entry.

04

Consensus review

Multiple contributions to the same feature are consolidated by weighted review. A trained typologist and a fluent community speaker both count, and they count differently.

Who we are looking for.

Linguists and typologists

Field linguists, descriptive grammarians, and typologists working on any language.

Translators

Professional translators who know where machine output goes wrong in your pair.

Native speakers

You do not need a linguistics degree. You need to know when a sentence is wrong.

Anthropologists and ethnographers

Cultural and pragmatic context that no grammar book records.

Universities and cultural institutions

Partner with us to host a Linguistics Excellence Centre for your region's languages.

Your work stays yours.

Named credit

Contributors are credited by name on the profiles they build, with a visible contribution history.

Co-authorship

Institutional partners have a co-authorship pathway on published research.

Terms up front

Contributed linguistic descriptions power the Mātr platform. Full contributor terms are provided with your application, before you contribute anything.

THE STANDARD

The profile is also the benchmark.

Because the framework can detect when an output breaks a language’s rules, the profile your community builds becomes the measure by which any AI system in that language can be judged. Universities and cultural institutions that host a Centre publish that standard in their own name.

Describe the language inside the framework.

Detect where any model violates it.

Publish the standard others are held to.

APPLY TO CONTRIBUTE

Tell us about yourself.