Discrete Cross-Modal Alignment Enables Zero-Shot Speech Translation

30799

Knowledge Triples

Source Entity	Relationship	Target Entity
Discrete Cross-Modal Alignment Method	employs	Vector Quantization Module Method
End to End Speech Translation Task	requires	Parallel ST Data Material
Vector Quantization Module Method	mapsTo	Shared Codebook OtherEntity
Zero Shot Speech Translation Task	proposes	Discrete Cross-Modal Alignment Method
Zero Shot Speech Translation Task	uses	Machine Translation Task