1 paper · 1 filter
Sangmin Lee, Woo-Jin Chung, Hong-Goo Kang
Building a universal multilingual automatic speech recognition (ASR) model that performs equitably across languages has long been a challenge due to its inherent difficulties. To a…