표 4. | Table 4. 트랜스포머(T), 컨포머(C), E-브랜치포머(E) 세 가지 인코더 모델별 각 데이터셋에 대한 음절 오류율(%) | Character error rate (%) for each dataset by encoder model: Transformer (T), Conformer (C), and E-Branchformer (E)

	fleurs-ko	kmsav	evalclean	evalother
T	4.54 (±0.03)	9.38 (±0.10)	6.83 (±0.03)	7.30 (±0.02)
C	4.30 (±0.04)	10.21 (±0.35)	6.54 (±0.06)	6.95 (±0.05)
E	4.11 (±0.11)	8.62 (±0.07)	6.17 (±0.00)	6.73 (±0.06)

The mean and standard deviation are shown for two models of the same structure trained with different initial values.