About the ALT-EDIC

The Alliance for Language Technologies (ALT-EDIC) is a European Digital Infrastructure Consortium supporting the development of Language Technologies. It will promote excellence in language technologies and contribute to the preservation of Europe’s linguistic diversity and cultural richness.

It was proposed by 10 Member States in December 2023 and officially created by a Commission decision in February 2024. It currently gathers 17 members and 8 observer Member States and regions. It is coordinated by France and its statutory seat is located in the historic site of the Château de Villers-Cotterêts.

It is currently hiring its core team in view of launching its operations in spring 2025.

The ALT-EDIC supports the development of language technologies

The Alliance for Language Technologies addresses the scarcity of European language data available for training Large Language Models (LLMs) and of benchmarking for such models.

This will lead to enhanced artificial intelligence (AI) solutions capable of understanding and generating language taking into account Europe’s linguistic diversity and cultural richness.

The ALT-EDIC’s mission focuses on the five following actions:

Action 1 - Data

ALT-EDIC will leverage on the Language Data Space and federate existing language and multimodal resources from the EU and Member States in all European, national and regional languages, including through the creation of strategic data such as for languages with few speakers (less than 10 million speakers) in which there are inherent limitations to training Large Language Models (LLMs).

Action 2 – Existing models

ALT-EDIC will create a repository of existing open-source language models for reuse by industrial actors and develop specific methods for fine-tuning especially for SMEs, and will provide evaluation, certification and normalisation methodologies with a particular focus on potential discrimination and bias introduced by Natural Language Processing (NLP) models.

Action 3 – Development of models

ALT-EDIC will act as a pool seed fund, bringing together public and private resources to launch and develop new Large Language Model projects and Foundation Models with multimodal capabilities, including through providing access to the necessary European High-Performance Computing.

Action 4 – Evaluation, certification and normalisation

ALT-EDIC will contribute to the development of evaluation methodologies with a particular focus on potential discrimination and bias introduced by NLP models as well for to provide dedicated support to institutions for investing in LTs.

Action 5 - Ecosystem

ALT-EDIC will act as an advisory point for public administrations as well as reaching the public through a cultural program based on artificial intelligence for languages and enabling LT end-users, who are also data producers, to take up the challenges of artificial intelligence and language technologies in a multilingual context and contribute to the enlightenment of the European citizen on the matter of artificial intelligence.