The Research Council has granted funding for a new project aimed at developing machine translation between Faroese and English. The project is led by Barbara Scalvini, assistant professor of computer science at the Faculty of Science and Technology at the University of the Faroe Islands. She is joined by Iben Nyholm Debess, a PhD Scholar at the Faculty of Faroese Language. Together, they are working at the Centre for Language Technology.
An exciting aspect of the project is that it requires participation from a broad range of Faroese language users. All language users can use the machine translator to translate texts between Faroese and English. Then, they can evaluate the translation and make any necessary changes to the translated text. These feedback contributions will be used to improve future translations through the continued development of the machine translation model. In this way, the average Faroese language user is able to directly influence which words and phrases are used in machine translations in the future, giving everyone a voice in the process. This approach is based on a linguistically democratic perspective and takes real-world usage into account.
Getting help from all language users is also crucial as large language models gain increasing influence, with their usage becoming more common in both everyday life and professional settings, including in the Faroe Islands. If Faroese-specific words, topics, and concepts are to have a presence on digital platforms in the future, it is necessary to provide machine translation models with these words, topics, and concepts. Contributions from Faroese speakers can therefore make a real difference.
To establish a translation platform, the first step is to develop the best possible machine translation model. This will be done by:
- Expanding bilingual datasets by leveraging the capabilities of large language models.
- Fine-tuning a multilingual model to translate between Faroese and English.
After that, the translation platform will be set up. Initially, it will assist users in translating texts, and later, it will help researchers further improve machine translation.
This project establishes a language technology infrastructure that not only benefits the Faroese language but can also be useful for other low-resource languages. A key aspect of the project is that everything will be open source. Translation models, high-quality bilingual datasets, and evaluation frameworks will be accessible to everyone—researchers, developers, companies, and the public—so that development can continue even after the project is completed. This lowers the barrier for small businesses or individuals to develop models for Faroese while also providing major tech companies with better opportunities to integrate Faroese language technology tools into their services.
In recent years, we have seen several large, foreign platforms offering translation to and from Faroese (such as Google Translate, ChatGPT, and Claude). This is encouraging, and many people use these services. However, it is crucial that Faroese speakers also take part in the development process. Having open models developed by us in the Faroe Islands allows us to retain control over language technology tools for Faroese and to cultivate local expertise in this rapidly evolving field.
The project is funded by the Research Council and the University of the Faroe Islands. Work on the project will begin immediately, and the translation platform is expected to be publicly accessible by October 2025. The project is set to be completed in the spring of 2026.
