Header

Search

Real-time Speech Transcription and Translation for Romansh

Romansh is a national language of Switzerland with several distinct varieties. In this project, you will extend our speech recognition systems for Romansh to support streaming, so that the transcription can be generated while the speaker is speaking.

You will have the opportunity to demonstrate your prototype live at Lia Rumantscha's Digital Day in Fall 2027 in Chur, Switzerland.

Optionally, you could then integrate speech-to-text translation using a streamed pipeline approach, or experiment with speech-to-speech translation using our speech synthesis system.

The project is supervised by Srikanth Madikeri and Jannis Vamvas.