no code implementations • 21 May 2023 • Oli Liu, Hao Tang, Sharon Goldwater
Self-supervised speech representations are known to encode both speaker and phonetic information, but how they are distributed in the high-dimensional space remains largely unexplored.