Leuven University (KU Leuven) and The Flemish Centre of Knowledge for Citizen Science (Scivil) have launched a platform to gather data on Flemish dialects and accents. Maarallee Platform’s aim is to gather data on our region’s rich variety of dialects and accent and to use the data gathered to train AI models.

Published:Fri 17 Oct. 10:42

Current Dutch-language speech recognition technology is mainly based on data from the Netherlands. This means that AI models have difficulty recognising the accents of Dutch-speaking Belgians.

Scivil’s AI and Education Project Officer Isaak Vandermaesen told the press agency Belga that “By recording your own voice on the maarallee.be platform you are helping to contribute to creating AI that is truly tailored to Flemish people.”

Anyone can record short sentences, which are automatically transcribed and then corrected by volunteers. This provides data that KU Leuven can use to develop a robust automatic speech recognition model for people across Flanders.

Those involved in the project say that the data collected will be both of scientific and social value. Speech recognition is playing an increasingly important role in education, the media, healthcare, industry and services. Examples of this can be found, in areas such as automatic subtitling, hands-free systems and voice-controlled tools for people that are living with disabilities.

Mr Vandermaesen said that “Without broad and diverse speech data, AI will remain technology for the few. With Maarallee.be, we are taking a step towards AI that can correctly understand every Fleming.”