From: Three-stage training and orthogonality regularization for spoken language recognition
Subset | Data composition [43] | Number of utterances | Duration (h) | Channel |
---|---|---|---|---|
ASR-train | AP16-OL7, AP17-OL3 | 50,071 | 71.42 | Mobile |
LID-train | ASR-train, AP17-OLR-test, AP18-OLR-test, AP19-OLR-test | 176,354 | 205.25 | Mobile |
dev | subset of ASR-train | 2992 | 4.53 | Mobile |
channel-test | AP20-OLR-channel-test | 11,848 | 17.84 | Cross-channel |
noisy-test | AP20-OLR-noisy-test | 9496 | 13.1 | Mobile |