Acoustic Model with Multiple Lexicon Types for Indonesian Speech Recognition

Figure 13

The %WER using 46,550 validated augmented vs. nonaugmented utterances.