AI at the university

Lately Barbara Scalvini has been popular in the Faroese media. The national radio recently did an interview with Barbara Scalvini about the University’s involvement in developing Faroese AI for machine translation and other purposes. The interview was broadcast in the news on Friday, 29 August.

In the interview, Barbara explained that several parts of this work are already available for use. Companies and institutions can access a programme that translates between English and Faroese. It is open source, free, and already available on Huggingface. It doesn’t take much to make the model work together with other systems—often just a few lines of code. You can see the models Barbara has uploaded to Huggingface by clicking here.

Barbara also emphasized,that an important part of the translation project is that Faroese people can contribute through a crowdsourcing platform, where they correct translations and thereby improve the models. This ensures that the AI doesn’t just learn the written language from books and newspapers, but also the everyday language you and I speak. This translation platform will be launched soon, and we will announce it when it becomes available.

In connection with Vísindavøka on Friday, the morning radio program is focusing on various research projects, including those led by Barbara. In the interview, she explains that Faroese AI does exist, but it is mainly language models with specific, limited functions – and not broad capabilities such as ChatGPT. Language models for Faroese include, for example, translation models, speech recognition models, text-to-speech models, and natural language understanding models. To bring these models closer to the user, one must build a service around them. An example of this is VoisIT. The greatest challenge in improving Faroese AI is that we lack sufficient data. She says that the Centre for Language Technology collaborates with Meta and Google, but that these large companies do not always invite Faroese into their services on their own initiative, and that we therefore have to knock on their doors to remind them that small languages also matter.

Barbara supervises two PhD students at the University who also work with AI. They are Dávid í Lág, who works on developing Faroese speech recognition,and Iben Nyholm Debess, who is developing an evaluation framework for Faroese, so that it will be possible to assess how different language technologies perform in Faroese.

Scroll to Top