Reinforcement Learning for Data-Efficient Code-Switched ASR

Ziwei Ye, Peter Vickers

Audio-language models can be prompted for code-switched speech, but their decoding is not optimized for code-switching and often fails at language boundaries. We propose a practica…