1 paper
Mingwei Fu, Zheshu Song, Zhikai Zhou +3
Current ASR models perform well on general benchmarks but still struggle with acoustically ambiguous words. Visual context can aid recognition in screen-based scenarios, yet existi…