The university and higher education sector
Several actors are relevant for adapting language models to Norwegian conditions.
The university and higher education sector
In the university and higher education sector, there are several initiatives that both develop pre-trained Norwegian language models and fine-tune international models, for example, the Norwegian Research Center for AI Innovation (NorwAI) at NTNU. The NORA consortium includes many collaboration partners, including UiO and Norwegian Computing Center.
The National Center for E-health Research developed a healthcare language model (NorDeClin-BERT).
Future research projects related to language models will play an important role in adaptation to Norwegian conditions. It is important that some of these initiatives include adaptation to the Norwegian health and care sector, i.e., healthcare language models.
The health and care sector
In the health and care sector, both Helse Vest IKT and Sørlandet Hospital have been involved in adapting language models to Norwegian conditions for research purposes. Helse Vest IKT is now also looking at specific use cases for the language model Clinical NorBERT.
Industry plays a very important role by fine-tuning language models and offering services based on language models, including Norwegian pre-trained models, but especially international models. Suppliers now offer services based on language models to, among others, general practitioners.
The National Library
The National Library is a central research institution that develops and adapts language models for Norwegian conditions. The National Library has developed and made available the language model NB-Whisper for transcription of audio recordings, and the model has been further developed by suppliers for the health and care sector.
The National Library will most likely have a special position in Norway when it comes to adapting language models to Norwegian conditions. The government proposes in the state budget for 2025 to allocate a total of 40 million kroner for training Norwegian and Sami language models (foundation models) for use in artificial intelligence. Of these, 20 million kroner will go to the National Library, which is tasked with training and making language models available, while 20 million kroner will go to financing computing power for training the models. The computing power will be delivered by the high-performance computing company Sigma2 AS, which is owned by Sikt – The Knowledge Sector's Service Provider [145]. The National Library's position could become very important for work with language models in the health and care sector as well, and close collaboration could be appropriate.