|
Project Description
|
Link
|
|---|---|
GPT-NL We ontwikkelen een taalmodel voor de Nederlandse taal en cultuur. Transparant, inclusief, en volgens Nederlandse en Europese waarden en zullen open over de beslissingen en afwegingen communiceren. Experts die aan dit project werken zijn Dominique Blok, Simon van de Fliert, Erik de Graaf, Eliza Hobo, Saskia Lensink, Jesse van Oort, en Daan Vos. |
Webpage |
EU-FarmBook EU-FarmBook is a Horizon Europe project that is working at regional, national, and European (EU) levels to build an Online Platform. Gathering and sharing agriculture and forestry knowledge. The main purpose of EU-FarmBook is to support knowledge exchange, by further developing, expanding and maintaining an easily accessible and user-friendly EU-wide Online Platform for practitioners in the agriculture and forestry sectors. Experts who have worked on this project are Daan Di Scala and Daan Vos |
Webpage |
TrustLLM TrustLLM is a European initiative to research the trustworthiness and factualness of language models, after which foundational language models are built to show it's capabilities. Several TNO experts are active within the project, for example focusing on the evaluation, architectual framework, and implementation in use-cases. TNO takes the lead in the use-cases, which entails ensuring the completion of each use-case. TNO experts focus on building Boekbot and Bijsluiterbot, two RAG implemented systems who are within the literary and medical domain. Experts who have worked on this project are Roos Bakker, Dominique Blok, Maaike de Boer, Daan Di Scala, Simon van de Fliert, Erik de Graaf, Saskia Lensink, and Stephan Raaijmakers. The project will span to 2026. |
Webpage |
Starlight Language models are dependent on an enormous amount of data. However, several domains have trouble in finding and collecting relevant and high-quality data, thus hamstringing the development of advanced systems. Within the HORIZON Starlight project, experts analysed the potential of open source language models to quickly synthesize new data points based on existing data, thus ensuring relevancy to the source domain. On top of this, evaluation frameworks were analysed to measure and ensure the quality of the synthetic datapoints. Experts who have worked on this project are Simon van de Fliert, Erik de Graaf. This project is set to continue until september 2025. |
Webpage |
Chatbot An asking party offers their clients a wealth of information on their website, however this information is spread thinly across hundreds of webpages. One of the feedback points of their clients was that the presented information is hard to find. We helped to collect the fragmented information and transform it into a usable dataset for Question and Answering. We then built a RAG (Retrieval Augmented Generator) and an evaluation framework that uses the built dataset as gold-standard answers. Moreover, we helped the asking party to prepare their system to scale to production. Experts who have worked on this project are Simon van de Fliert, Romy van Drie, and Eliza Hobo. This project continues to 2025. |