SaameAI – Speech recognition for the Saami languages
Project
SaameAI – Speech recognition for the Saami languages is a co-project of Aalto University and the University of Lapland. The aim of the project is to develop an automatic speech recognition (ASR) system for the Saami languages. First, the system is developed for North Saami, but in the future we will also expand for other Saami languages. The project is estimated to last for three years, and it has received funding from the Finnish Cultural Foundation. The University of Helsinki is collaborating with the project.
Speech recognition tools for the Saami languages
There is a growing demand for different technological solutions and tools among the Saami languages. An automatic speech recognition, or speech-to-text, tool can be beneficial, for instance, in fields such as public services, education, and research. Transcribing speech to text manually is a slow and time-consuming job that often requires specialized skills or training. Automatic tools designed for this task will spare time and resources.
Developing a speech recognition system for the Saami languages is also a question of linguistic equity. Language technology solutions have been developed for only a few of the world’s languages, which also happen to be the ones with the most resources – both material and human – available. While smaller and endangered languages do not have the same resources, the benefit that different language technology tools may offer for these languages is significant. Besides the time and resources saved by automating some of the work stages, language technology tools may offer possibilities or support to use languages in new and different settings.
Updates and progress
A speech recognition model has been built at Aalto University using data provided by the Finnish National Audiovisual Institute (KAVI) and the Norwegian Saami Parliament. A demo version of the model is publicly available at HuggingFace .
Anyone may test the tool on this website, but please be aware that we do not recommend uploading any private or sensitive data on this page, since we cannot ensure information security on this page. An ASR program that runs in an information secure environment is currently under development at Aalto University.
The accuracy of the model will be further improved in the future by training it with more data. The training and fine-tuning require audio material with corresponding text.
Contact
Project manager
Mikko Kurimo
Aalto-university, Professor, Information and communication engineering, speech recognition
E-mail mikko.kurimo@aalto.fi
Phone +358503476221
Project partner
Pigga Keskitalo
University of Lapland
Professor, Education sciences
E-mail
pigga.keskitalo@ulapland.fi
Doctoral researcher
Kristiina Ojala
University of Helsinki
Doctoral researcher, Language studies program
E-mail kristiina.ojala@helsinki.fi
Phone +358440674616
Partners and collaborators
Mikko Kurimo (Aalto-university), project leader
Pigga Keskitalo (University of Lapland), project partner
Kristiina Ojala (University of Helsinki), doctoral researcher
Yaroslav Getman (Aalto-university), doctoral researcher
Riho Grünthal (University of Helsinki), PhD supervisor