NORDTRANS - Technology for automatic speech transcription in selected Nordic languages

Project facts

Project promoter:
NEWTON Technologies, a.s.(CZ)
Project Number:
CZ-RESEARCH-0022
Status:
Completed
Final project cost:
€1,408,795
Donor Project Partners:
Norwegian University of Science and Technology(NO)
Other Project Partners
Technical University of Liberec(CZ)
Programme:

More information

Description

The main goal of the project is to improve the state-of-the-art quality and usability of the automatic speech recognition (ASR) technology for Swedish and Norwegian. In contrast to existing solutions, the newly developed technology will operate with high accuracy in all various applications including online broadcast monitoring, such as TV, internet podcasts, etc., transcription of speeches in parliaments and similar public institutions, as well as spoken archive mining. The incorporation of the developed technology into the existing multilingual speech processing solutions of NEWTON Technologies will open new market and cooperation opportunities for Newton in Northern Europe and, at the same time, bring new or better services based on ASR closer to people living in this part of Europe.

Summary of project results

The NORDTRANS project aimed to create accurate automatic speech recognition (ASR) systems for the transcription of the Nordic languages, particularly Norwegian and Swedish. This included handling the two written standards of Norwegian (Bokmål and Nynorsk).

The project developed ASR systems for Norwegian and Swedish using end-to-end models. They achieved globally competitive transcription technology in three Nordic languages, surpassing benchmarks of commercially available systems in some cases. They also created Danish ASR and made the software modules available on the Beey platform.
 

The project developed accurate and robust ASR systems for Norwegian and Swedish, surpassing the performance of existing commercial systems in some cases. This research also contributed to a better understanding of the linguistic and acoustic characteristics of Norwegian and Swedish. The developed technology benefits the general public, broadcasters, parliament, and researchers by providing accurate and efficient speech recognition tools. The results have the potential to significantly impact various sectors and contribute to the development of language technology in the Nordic region.
 

Summary of bilateral results

The project would not have been successful without bilateral cooperation, as it is a primarily Norwegian dialect transcription in which all partners cooperated in mutual research, and have enriched their know-how. Further collaboration is planned.

Information on the projects funded by the EEA and Norway Grants is provided by the Programme and Fund Operators in the Beneficiary States, who are responsible for the completeness and accuracy of this information.