Concetti Chiave
Self-supervised fine-tuning method SCORE improves content representations efficiently.
Statistiche
SCORE fine-tuned HuBERT outperforms vanilla HuBERT with relative improvements of 1.09%, 3.58%, and 12.65%.
SCORE provides competitive results with SPIN using only 1/3 of the processed speech compared to SPIN.
Citazioni
"SCORE fine-tuned models outperform vanilla HuBERT and WavLM on the SUPERB benchmark."
"SCORE requires the least amount of processed speech compared to other SSFT methods."