2 papers
cs.CL2025
The Limits of Data Scaling: Sub-token Utilization and Acoustic Saturation in Multilingual ASR
Siyu Liang, Nicolas Ballier, Gina-Anne Levow +1
How much audio is needed to fully observe a multilingual ASR model's learned sub-token inventory across languages, and does data disparity in multilingual pre-training affect how t…
cs.CL2025
Beyond WER: Probing Whisper's Sub-token Decoder Across Diverse Language Resource Levels
Siyu Liang, Nicolas Ballier, Gina-Anne Levow +1
While large multilingual automatic speech recognition (ASR) models achieve remarkable performance, the internal mechanisms of the end-to-end pipeline, particularly concerning fairn…