Table 3.
Phone tokens for correct and incorrect pronunciations on different datasets.
Data Collection | Phones in Total | Phones with Pronunciation Error | Pronunciation Error Rate % |
---|---|---|---|
Training Set PSC-Train-1000 |
408,000 | 50,616 | 12.41% |
Develop Set PSC-Develop-87 |
35,496 | 4432 | 12.49% |
Test Set PSC-Test-89 |
36,312 | 4544 | 12.51% |
Total | 479,808 | 59,592 | 12.42% |