Topsector ICT has awarded funding to two European research projects involving TNO and various Dutch businesses and knowledge institutions. The projects are:
- Large Language Models for the European Union (LLMs4EU), an initiative by the Alliance for Language Technologies (ALT-EDIC), running from 1 April 2025 to 31 March 2028.
- MediSpeech, an international partnership seeking to improve administrative efficiency in health care settings using advanced speech recognition and AI.
According to the director of Topsector ICT, Frits Grotenhuis, the projects are seamlessly aligned with the Knowledge and Innovation Agenda (KIA) Digitalisation. This Agenda, initiated and coordinated by Topsector ICT, is based on three pillars: innovation in digital and information technologies (DITs), innovation with DITS and reflection on DITs. ‘Topsector ICT is currently coordinating the development of an Action Agenda for AI/Data, working in close collaboration with relevant parties. AI/Data is one of the ten priority key enabling technologies in the Dutch government’s National Technology Strategy (NTS),’ explains Grotenhuis.
For the LLMs4EU project, Topsector ICT has earmarked €929,046 in public-private partnership innovation (PPSi) funding aimed at digital innovation. For MediSpeech, Topsector ICT has earmarked €350,734 in public-private funding.
More about LLMs4EU
Large Language Models (LLMs) are a major development in Artificial Intelligence (AI), representing a fundamental shift in how machines understand, generate and interact with human language. They have gained recognition among the wider public, who have been able to experience the potential of AI in this domain.
The total project budget for the European consortium amounts to almost 40 million euros, provided by the EU’s Digital Europe programme. A total of 2.7 million euros is earmarked for the Dutch part of the research, for which TNO is acting as secretary.
LLMs4EU is an ambitious European initiative to develop open-source LLMs for all official EU languages, including minority languages, to ensure that generative AI drives digital inclusion in the EU. This ties in with the EU’s aim of countering the current market dominance of US and Chinese LLMs.
All models, training data and software in the project are open source, guaranteeing transparency and modifiability. The models are also being developed in according with European legislation and regulations such as the EU AI Act and the General Data Protection Regulation (GDPR).
Major Dutch contribution
The consortium brings together over 20 European countries, including the Netherlands. Participating in the project in the Netherlands are TNO and various Dutch companies, such as KPN, as well as a range of knowledge institutions, including SURF, Eindhoven University of Technology (TU/e) and the University of Twente (UT). The Dutch researchers will focus on collecting data, establishing the correct infrastructure and setting up a virtual evaluation centre for LLMs.
The Netherlands’ participation in this European LLM development project is helping to reduce reliance on non-European technologies while ensuring that Dutch and other smaller languages are effectively supported in AI applications. The project’s open-source nature means that Dutch businesses and research institutions can enhance the models independently and develop innovation applications for various sectors, including telecoms, tourism, energy, public services and health care.
Find out more about the project
Find out more about the consortium
More about MediSpeech
MediSpeech is another innovative research project for which Topsector ICT has earmarked funding. This international partnership aims to improve administrative efficiency in health care settings using advanced speech recognition and AI. MediSpeech seeks to achieve automated medical reporting by:
- Using AI-powered speech recognition technology specifically designed for clinical use;
- Integrating terminologies such as MesH, SNOMED, LOINC and ICD to ensure data consistency and interoperability;
- Using advanced natural language processing (NLP) and LLMs to automate medical reporting.
By combining these technologies, MediSpeech seeks to create an open digital ecosystem that improves medical reporting efficiency and enhances user experience for patients and health care professionals alike.
Role of Dutch organisations
The project will run from January 2025 to December 2027. TNO is one of the 22 participating partners in the project, joining others from five countries. As a research organisation that specialises in transforming scientific knowledge into practical applications for business, public administrations and wider society, TNO is playing a crucial role in MediSpeech by contributing its expertise in digital health and AI. The MediSpeech project is led by MEDrecord BV, a fellow Dutch company, and involves other Dutch partners such as AMC Amsterdam, HealthTalk.ai, ORTEC, Radboud University and the University of Twenty (UT). Parties from Finland, Portugal, Türkiye and the UK are also on board.
Expected impact
MediSpeech has the potential to revolutionise the way in which medical information is recorded and shared. Automating and standardising administrative processes will enable health care providers to focus more on patient care, reduce the likelihood of errors and improve interoperability between different health care systems.
This project is funded by the ITEA 4 programme, part of the Eureka RD&I cluster on software innovation in the EU.