Table 2.
Contains the specifics of the Common Voice and SpeechOcean datasets. The ‘training’ set trains the ASR model, while the ‘testing’ set tests its accuracy.
| Dataset | Total Words | Total Unique Words | Total Duration (h) | Total Utterances | Average Duration (s) | Average Word per Utterances |
|---|---|---|---|---|---|---|
| Common Voice (training) Common Voice (testing) SpeechOcean |
497,326 | 78,604 | 127.41 | 117,526 | 4.27 | 4 |
| 23,450 | 14,108 | 11.69 | 8840 | 4.23 | 5 | |
| 406,904 | 50,256 | 79.95 | 63,284 | 5.93 | 8 | |
| Common Voice (training) + SpeechOcean | 904,230 | 128,860 | 207.36 | 180,810 | 4.80 | 6 |