Language models are a type of artificial intelligence that is capable of understanding and generating human language. For example, they can be used to develop chatbots that are able to understand and respond to users. However, there is a need to develop transparent, Danish language models to solve limited tasks, where an understanding of Danish norms and values is complied with by GDPR and the AI Regulation. With a new grant from the Ministry of Digitalisation, it will now be possible to establish a secure platform for the development of Danish language models.

Aarhus University and the University of Southern Denmark, together with the Alexandra Institute and the University of Copenhagen, have received DKK 30,7 million from the Ministry of Digitalisation to establish a secure platform for the development of transparent Danish language models.

From Aarhus University, Kristoffer Laigaard Nielbo, professor at the Center for Humanities Computing, will head the project.

"By taking Danish cultural heritage as a starting point and by making culturally adapted adjustments to already existing models, we want to create language technology that reflects and respects the complexity of Danish society – then and now," he says.

Language models will meet the highest standards

The focal point of the project is the establishment of a secure platform for research and development of language models that Danish companies and authorities must be able to access. The platform will also be able to develop smaller, specialised language models that all meet the highest standards for data integrity and documentation of safe AI use.

In the long term, the models will be made freely available to companies and authorities, which will then have the opportunity to use and further develop Danish language models for their own usage needs.

"There is an obligation to openness and transparency in the research and development work. In other words, we want to contribute not only with the best models, but also with tools and documentation that enable further development and reproduction of our work. We thus strive to bridge the digital language barrier and to mature the norms for how we culturally sound develop AI," says Kristoffer Laigaard Nielbo.

Publicity based on a press release from the Ministry of Digital Affairs and Alexandra Instituttet

About Center for Humanities Computing
Center Humanities Computing is involved in the work with Danish Foundation Models (DFM) - a collaborative project that develops advanced language models specifically for Danish. DFM originates from a vision that community and inclusion must be the guiding principles for the development of Danish language technology.

The project is run by the Alexandra Institute, the University of Copenhagen, the University of Southern Denmark and Aarhus University. Kristoffer Lajgaard Nielbo is the project manager from Aarhus University, Peter Schneider-Kamp from the University of Southern Denmark, Jens Kaas Benner from the Alexandra Institute and Bolette Sandford Pedersen from the University of Copenhagen.

Together, they work to create models that support the Danish language and culture, and that can be used widely in both the public and private sectors. DFM has an extensive partnership with the Danish Language Model Consortium, which includes both public and private actors. Among the participants are Aarhus Municipality, ATP, Falck, Salling Group, Topdanmark and Visma Enterprise.

Contact

Kristoffer Laigaard Nielbo, Professor
Centre for Humanities Computing
School of Culture and Society
Aarhus University
Mail: kln@cas.au.dk
Mobile: +45 2683 2608