AISHELL-3 北京希尔贝壳科技有限公司 Beijing Shell Shell Technology Co.,Ltd 10/18/2020 1. AISHELL-3 Speech Data - Sampling Rate : 44.1kHz - Sample Format : 16bit - Environment : Quiet indoor - Speech Data Type : PCM - Channel Number : 1 - Recording Equipment : High fidelity microphone - Sentences : 88035 utterances - Speaker : 218 speakers (43 male and 175 female) 2. Data Structure │ README.txt (readme) │ ChangeLog (Change Information) │ phone_set.txt (phone Information) │ spk_info.txt (Speaker Information) └─ test (Test Data File) └─ train (Train Data File) │├─content.txt (Transcript Content) │├─prosody_label_train-set.txt (Prosody Lable) │├─wav (Audio Data File) │├─SSB005 (Speaker ID File) ││ ││ ID2166W0001.wav (Audio) 4. System AISHELL-3 is a large-scale and high-fidelity multi-speaker Mandarin speech corpus which could be used to train multi-speaker Text-to-Speech (TTS) systems. You can download data set from: http://www.aishelltech.com/aishell_3. The baseline system code and generated samples are available online form: https://sos1sos2sixteen.github.io/aishell3/.