no code implementations • 28 Mar 2022 • Shashank Subramanya, Jan Niehues
Based on a technique to adapt end-to-end monolingual models, we investigate multilingual models and different architectures (end-to-end and cascade) on the ability to perform online speech translation.