Homophone Disambiguation Reveals Patterns of Context Mixing in Speech Transformers

Open in new window