Table 1.
Statistics and experimental setting of the CMaiSpeech and NER data sets.
| Methods | CMaiSpeecha data set | NERb data set | CMaiSpeech data set | |||||||||
|
|
Training | Validation | Training | Validation | Testing | |||||||
| Wave statistics | ||||||||||||
|
|
Duration, hours | 105.5334 | 1.6572 | 114.2523 | 12.4390 | 1.8093 | ||||||
|
|
Mean duration, seconds | 14.2634 | 18.9395 | 21.6751 | 21.2028 | 17.5566 | ||||||
|
|
Utterances, n | 26,636 | 315 | 18,976 | 2112 | 371 | ||||||
| Word statistics, n | ||||||||||||
|
|
English words | 136,196 | 2342 | 682 | 150 | 2427 | ||||||
|
|
Chinese words | 787,701 | 12,115 | 1,420,101 | 206,909 | 14,820 | ||||||
|
|
Total words | 923,897 | 14,457 | 1,420,784 | 207,059 | 17,247 | ||||||
| Code-switching level | ||||||||||||
|
|
Language entropy | 0.6033 | 0.6391 | 0.006 | 0.0086 | 0.5861 | ||||||
|
|
I-indexc, mean (SD) | 0.2006 (0.1529) | 0.1896 (0.16) | 0.0006 (0.0052) | 0.0011 (0.008) | 0.1411 (0.1372) | ||||||
aCMaiSpeech: China Medical University Hospital Artificial Intelligence Speech.
bNER: National Education Radio.
cI-index: integration index (ie, the probability of switching).