Skip to main content
Aller à la page d’accueil de la Commission européenne (s’ouvre dans une nouvelle fenêtre)
français fr
CORDIS - Résultats de la recherche de l’UE
CORDIS

Automating language documentation

Objectif

Almost half of the world’s 7000 languages are endangered. Linguists try to combat this loss of human heritage by documenting languages. They do this by recording words, sentences and stories in the original language, accompanied by transcriptions, translations and morpheme-by-morpheme analyses. The current documentation rate is too low to capture all the world's language before they disappear.

The labor-intensive process of documentation can be sped up with the help of machine learning, and in particularly large language models (LLMs). In this project, I will research how exactly LLMs can support linguists in these tasks, and how (field) linguists can adapt their data collection methods so that they profit maximally from automation.

I have created a task list with the most common language documentation tasks, and have ranked them according to time consumption, ease for a language model to tackle, and importance for language documentation. Starting with the highest ranking task, I and my team will fine-tune LLMs with a limited amount of existing data from my Uruangnirin language corpus, perform the task, and evaluate the results in terms of quality and time consumption compared to a human (linguist) baseline. To test for scalability, we will also implement the results live on two previously undocumented languages with different language. A PhD student and a postdoc will work with data from these languages, applying the tools and methods we develop based on the Uruangnirin data.

Scripts and guidelines for applying the models on other datasets will be published for each task. At the end of the project, I will publish a set of guidelines for linguistic fieldworkers that want to optimize data-gathering for language model processing. I will also organize workshops to help other fieldworkers getting started with using our tools and methods.

Champ scientifique (EuroSciVoc)

Ce projet n'a pas encore été classé par EuroSciVoc.
Proposez les domaines scientifiques qui vous semblent les plus pertinents et aidez-nous à améliorer notre service de classification.

Mots‑clés

Programme(s)

Thème(s)

Régime de financement

HORIZON-ERC - HORIZON ERC Grants

Voir tous les projets financés dans le cadre de ce programme de financement

Appel à propositions

(s’ouvre dans une nouvelle fenêtre) ERC-2026-STG

Voir tous les projets financés au titre de cet appel

Institution d’accueil

UPPSALA UNIVERSITET
Contribution nette de l'UE
€ 1 478 450,00
Adresse
VON KRAEMERS ALLE 4
751 05 Uppsala
Suède
Région
Östra Sverige Östra Mellansverige Uppsala län
Type d’activité
Higher or Secondary Education Establishments
Liens
Coût total
€ 1 478 450,00

Bénéficiaires (1)