Strategically investing in applied language technology

Thursday, March 4, 2021

AI Sweden’s new strategic program Applied Language Technology will strengthen Swedish competitiveness and ensure that we develop powerful Swedish language models that benefit our society. Magnus Sahlgren, Head of Natural Language Processing at RISE, will be the Research Lead and work together with Love Börjeson, Director of KB Lab at the National Library of Sweden, who will be the Data and Infrastructure Lead. They shared their view on the future of Swedish language models.

“The Swedish language is spoken by over 10 million people and so it is not a tiny language. Still, it pales in comparison to the English or Chinese speaking populations, which also drive AI and digital development”, Love Börjeson explains.

Applied language models lay the foundation for powerful AI solutions such as voice assistants at home or reliable automation of social services. But the most powerful and popular language models are currently based on English and Chinese. This points to the need for developing robust Swedish language models to ensure that Sweden strengthens its competitiveness and can keep up with delivering high quality welfare to its citizens.

To accelerate innovation, development of applications and implementation in this area, Applied Language Technology therefore constitutes one of AI Sweden’s new strategic programs. It is fueled by long-term investments and a number of underlying projects and use cases.

"This is great news for all of Sweden - in order to stay competitive, we need this large investment in Applied Language Technology. During the next few years we foresee a steady focus on text-based data, but multimodal data - video, images, sound - are catching up in importance. The strategic program at AI Sweden allows us to explore this, both across sectors and in much more depth. I envision that this will have a huge impact on both AI research and application." says Magnus Sahlgren, Research Lead Applied Language Technology at AI Sweden

Moreover, there are exciting opportunities to already today collaborate with our neighboring countries and languages. The recently launched pilot study for federated language models is an important step towards leveraging the Scandinavian languages and building a common model that could compete with the dominant models in English and Chinese.

"The National Library have very large collections that are an untapped potential for building internationally competitive language models. Partnering with AI Sweden allows us at KB Lab to exist in new contexts which pushes us to be even better at what we do, and the collaboration also with other experts works seamlessly. The research that we do has the potential to impact society in groundbreaking ways. Sharing this with and via AI Sweden increases the application of the models in real-life contexts. That is how we see completely new use case areas pop up!” says Love Börjeson, Data and Infrastructure Lead Applied Language Technology at AI Sweden and Director of KB Lab at the National Library of Sweden