On Thursday, March 14, 2024, a significant innovation in speech technology was presented, led by a team of Czech and Norwegian experts. The achievement was presented at a press conference in Trondheim, Norway, with an online broadcast. It is the successful completion of an international partnership between NEWTON Technologies, the Technical University of Liberec (TUL) and the Norwegian University of Science and Technology in Trondheim (NTNU).

The main joint project output of this international team is cutting edge speech-to-text transcription technology in Norwegian, Swedish and Danish. According to the final testing, the modules from the NORDTRANS project surpassed the world competition in transcription accuracy in all three Nordic languages.

"On average, our models significantly outperform the competing technologies; by a few percentage points in Danish and Swedish, while in Norwegian the difference in average results is smaller but still significant," explains Petr Červa, TUL team leader. 

NEWTON Technologies has integrated these cutting-edge, high-accuracy language modules into its Beey platform. This online tool gives users around the world easy access to the latest technologies for a variety of uses, from video captioning, meeting minutes and conference transcriptions to automatic summarization and other AI-powered processing.

Online broadcast of the press conference from March 14, 2024

"Large Language Models (LLMs), which are currently being talked about everywhere, are based primarily on the written form of languages. However, most digital content is in audio or audiovisual format: up to two-thirds of all internet traffic is video. When you convert speech from audiovisual sources into written text, you can also apply the power of artificial intelligence to this content," adds Alessandro Cederle, Business Development Manager at NEWTON Technologies, on the impact of speech-to-text technology on the current wave of AI in the digital world.

The project "NORDTRANS - Technology for Automatic Speech Transcription in Selected Nordic Languages" was funded by a grant of EUR 1,244,000 from the Norwegian Funds through the Research Programme (KAPPA).

For more information and materials from the conference see https://beey.io/trondheim.