binxia commited on
Commit
1f59198
·
verified ·
1 Parent(s): 9404c72

Upload 4 files

Browse files
.gitattributes CHANGED
@@ -57,3 +57,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
57
  # Video files - compressed
58
  *.mp4 filter=lfs diff=lfs merge=lfs -text
59
  *.webm filter=lfs diff=lfs merge=lfs -text
 
 
 
57
  # Video files - compressed
58
  *.mp4 filter=lfs diff=lfs merge=lfs -text
59
  *.webm filter=lfs diff=lfs merge=lfs -text
60
+ AISHELL2-Data-Specification\[EN\].docx filter=lfs diff=lfs merge=lfs -text
61
+ AISHELL2-Data-Specification\[ZH\].docx filter=lfs diff=lfs merge=lfs -text
AISHELL-2_README.txt ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ File : README
2
+
3
+
4
+
5
+ AISHELL-2 Speech Data
6
+
7
+ 北京希尔贝壳科技有限公司
8
+ Beijing Shell Shell Technology Co.,Ltd
9
+ 03/10/2018
10
+
11
+ 1. AISHELL-2 Speech Data
12
+
13
+ - Sampling Rate : 16kHz
14
+ - Sample Format : 16bit
15
+ - Environment : Indoor
16
+ - Speech Data Type : PCM
17
+ - Channel Number : 1
18
+ - Recording Equipment : iOS
19
+ - Time : 1000 hours of speech data (around 1 million utterances)
20
+ - Speaker : 1991 speakers (845 male and 1146 female)
21
+
22
+
23
+ 2. 数据目录结构
24
+ │ AISHELL2-Data-Specification .docx (Corpus Information)
25
+ │ README.txt (readme)
26
+ │ ChangeLog (Change Information)
27
+ └─ data (Data File)
28
+ │├─wav.list (Audio list)
29
+ │├─trans.txt (Transcript Content)
30
+ │├─spk_info.txt (Speaker Information)
31
+ │├─wav (Audio Data File)
32
+ │├─D2166 (Speaker ID File)
33
+ ││ ││ ID2166W0001.wav (Audio)
34
+
35
+ 3. Evaluation Data:
36
+ Currently we open-sourced AISHELL2-2018A-EVAL
37
+ - dev: 2500 utterances from 5 speaker
38
+ - test: 5000 utterances from 10 speakers
39
+ You can download above evaluation set from: https://www.aishelltech.com/aishell_2018_eval
40
+
41
+ 4. RECIPE
42
+ Based on Kaldi standard system, AISHELL-2 provides a self-contained Mandarin ASR recipe
43
+ - a word segmentation module, which is a must-have component for Chinese ASR systems
44
+ - an open-sourced Mandarin lexicon(DaCiDian)
45
+ - a simplified GMM training recipe
46
+ - 80-dim FBank without pitch as NN input feature
47
+ - slightly different data augmentation setup(tempo perturbation, max-volume perturbation)
48
+ - acoustic channel adaptation recipe(AM fine-tuning)
49
+ - a real-time streaming ASR demo via laptop microphone (tested on MacOS)
50
+ You can download recipe set from: http://www.aishelltech.com/aishell_2
AISHELL2-Data-Specification[EN].docx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f47a2edbfde6589757a6f3597ac88bf88ce9f19efa11347dfb0505b0cd295ee
3
+ size 145528
AISHELL2-Data-Specification[ZH].docx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c3a82459705d11350027fb791ea0387c3fe07f0ac08e04c2a6f8ca204ec72fb
3
+ size 154677
ChangeLog ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ ======================================================
2
+
3
+ AISHELL-2 Corpus ChangeLog
4
+ Version : 1.0
5
+ Date : 03/10/2018
6
+
7
+ ======================================================
8
+ ### v 1.0 (03/10/2018)
9
+ - AISHELL-2 is by far the largest open-source speech corpus available for Mandarin ASR research.