EU funding for research on LLMs

10/8/2024
Reading time 1 min.

Alexander Fraser, Professor of Data Analytics & Statistics at TUM Campus Heilbronn, has been awarded a prestigious ERC Advanced Grant by the European Research Council. His research project focuses on improving the integration of lesser-spoken languages into Large Language Models (LLMs), the AI-driven technology behind ChatGPT and similar chatbot platforms.

ERC Advanced Grants are awarded annually across several categories and are designated for leading researchers who have demonstrated exceptional accomplishments over the past decade. Each grant is worth up to €2.5 million. Since 2007, researchers at the Technical University of Munich have secured 228 ERC-funded projects, including 111 Starting Grants, 48 Consolidator Grants, 43 Advanced Grants, 3 Synergy Grants, and 23 Proof of Concept Grants.

EPICAL

In recent years, intelligent chatbots such as ChatGPT have revolutionized content generation. The quality of the results depends on how much content is available on the internet in the respective language. This means that languages such as English dominate. However, over 7000 languages are spoken worldwide. But integrating so-called low-resource languages into language models is difficult. Prof Alexander Fraser is tackling this challenge in his EPICAL project. The project’s goals include the automatic generation of error-free texts for low-resource languages and the development of better methods for language adaptation by combining related languages. Thus, the project thus helps to appreciate the role of languages with few digital resources and, on the other hand, to promote the solutions for machine learning problems with a small data basis in other research areas.

Alexander Fraser is Professor of Data Analytics & Statistics at TUM Campus Heilbronn. He is also a Principal Investigator of the Munich Centre for Machine Learning and a core member of the Munich Data Science Institute.

Further information and links