1 paper
Hanlin Zhang, Daxin Tan, Dehua Tao +3
Semantic speech tokens should preserve linguistic content while suppressing speaker- and duration-dependent variation inherited from acoustic inputs. We propose Iterative Semantic…