API
< Back to all clusters
[TECHNOLOGY] · Spain · 2 sources

started · updated

Conselh Generau launches campaign for Aranese AI speech recognition

The Conselh Generau, through its Departments of Language and Innovation, has launched a campaign to enable artificial intelligence to speak Aranese. As part of the LINGUATEC-IA project, the initiative aims to develop an Automatic Speech Recognition (ASR) system and a voice synthesizer for the language.

To achieve this, the government requires at least 175 hours of transcribed audio featuring a diverse range of speakers. While 8 hours were recorded during a 2022 voice marathon, an additional 167 hours are needed. The government is calling on citizens to participate by recording audio to ensure the model captures a wide variety of accents, genders, ages, and regional variations from the different areas of the Val d’Aran, including Naut Aran, Mijaran, and Baish Aran.

To maintain data diversity, no single speaker is permitted to contribute more than 2-3% of the total recordings, limiting individual contributions to approximately 1.5 hours. The Conselh Generau will provide a professional recording studio and specific texts to ensure a standardized format for the project, which is being carried out in collaboration with Lo Congrès permanent dera Lengua occitana.

Entities

Conselh Generau d'Aran · Lo Congrès permanent dera Lengua occitana · Val d’Aran