Vicomtech achieves the second place in the 2020 Albayzín Challenge for Spanish transcription systems using less training data

The Vicomtech Speech and Natural Language Technologies team has obtained the second place in the 2020 Albayzín Challenge, in the audio-to-text transcription modality.

This modality has had the participation of 7 national and international research groups and Vicomtech has obtained the second place with a speech recognition system trained with 5 times less data than the winning system, which represents a great technological milestone for the team. .

The Albayzín challenge has consisted in applying a Spanish transcription system in the complex domain of RTVE's television content, having evaluated a total of 55 hours of different types of programs. Vicomtech has solved the challenge by applying state-of-the-art speech recognition technology based on Deep Learning techniques through the proprietary technology asset TRANSKIT.

This second edition of the technological challenge organized by RTVE and Zaragoza University for research groups and companies in the field of speech technologies, artificial vision and natural language processing coincides with the IberSPEECH 2020 event, which will take place in Valladolid from 24th to 26th March.