Hugging Face's logo

Datasets:
bofenghuang
/
stt-pseudo-labeled-whisper-large-v3-multilingual

License:
Dataset card Files Community
1
bofenghuang commited on
Commit
a02f7fa
·
verified ·
1 Parent(s): 8f1b7bb

Add files using upload-large-folder tool

Browse files
Files changed (4) hide show
  1. README.md +234 -0
  2. __pycache__/meta.cpython-310.pyc +0 -0
  3. dur.txt +138 -0
  4. get_stats.py +84 -0
README.md ADDED
@@ -0,0 +1,234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-3.0
3
+ ---
4
+
5
+ This collection includes over 189,000 hours of speech-to-text data in seven languages: English, French, Spanish, Portuguese, Italian, German, and Dutch
6
+
7
+ All segments were initially sorted by their IDs (timestamps). Adjacent segments from the same source were concatenated into 30-second chunks before being decoded using Whisper-Large-V3. The only exception was Common Voice, where segments were decoded individually before concatenation.
8
+
9
+ In total, over 288,000 hours of audio data were collected and processed. This dataset retains only segments with a word error rate (WER) below 20% after normalization. Users can apply stricter filters if needed for their specific use cases.
10
+
11
+ ## Usage
12
+
13
+ ```python
14
+ from datasets import load_dataset
15
+ from pprint import pprint
16
+
17
+ dataset = load_dataset(
18
+ "bofenghuang/stt-pseudo-labeled-whisper-large-v3-multilingual",
19
+ "en-ami-ihm", # can also select other subsets
20
+ split="train",
21
+ trust_remote_code=True,
22
+ )
23
+
24
+ pprint(dataset[0])
25
+ # {'audio': {'array': array([-6.10351562e-05, -1.22070312e-04, -1.83105469e-04, ...,
26
+ # 6.40869141e-04, 6.40869141e-04, 6.40869141e-04]),
27
+ # 'path': None,
28
+ # 'sampling_rate': 16000},
29
+ # 'audio_filepath': '/home/bhuang/.cache/huggingface/hub/datasets--bofenghuang--stt-pseudo-labeled-whisper-large-v3-multilingual/snapshots/8f1b7bbae8f1b657d1a95b66dba9d2e7b3b86665//distil-whisper/ami-ihm/ihm/train_concatenated/EN2001a.zip:62624564:958444',
30
+ # 'duration': 29.950000762939453,
31
+ # 'prev_text': "OKAY DOES ANYONE WANT TO SEE UH STEVE'S FEEDBACK FROM THE "
32
+ # "SPECIFICATION RIGHT NOT REALLY UM JUST WHAT HE'S TALKING ABOUT "
33
+ # 'LIKE DUPLICATION OF EFFORT AND LIKE DUPLICATION OF EFFORT AND '
34
+ # 'STUFF AND UM YEAH HE WAS SAYING THAT WE SHOULD MAYBE UH THINK '
35
+ # 'ABOUT HAVING A PROTOTYPE FOR WEEK SIX WHICH IS NEXT WEEK YEAH '
36
+ # 'SO WE SHOULD PROBABLY PRIORITIZE OUR PACKAGES MM YEAH YEAH HMM',
37
+ # 'prev_whisper_transcript': "<|0.00|> Does anyone want to see Steve's feedback "
38
+ # 'from the specification?<|4.80|><|4.80|> Not '
39
+ # 'really, just what he was talking about, like '
40
+ # 'duplication of effort and '
41
+ # 'stuff.<|11.20|><|11.20|> And saying that we '
42
+ # 'should maybe think about having a prototype for '
43
+ # 'week six, which is next week.<|21.00|><|21.00|> '
44
+ # 'So we should probably prioritise our '
45
+ # 'packages.<|28.34|>',
46
+ # 'text': 'HAS HAS ANYONE ACTUALLY LOOKED AT THE JAVA CODE FOR THE HUH HMM YEAH '
47
+ # "I THINK SO YEAH I I DON'T KNOW ABOUT THE SEARCH FUNCTIONALITY THAT "
48
+ # "MIGHT BE ONLINE DEPENDS HOW IT'S GONNA WORK YEAH MM-HMM YEAH THAT "
49
+ # 'MAKES SENSE HMM HMM YEAH YOU JUST CONCATENATE THEM TOGETHER HMM YEAH '
50
+ # 'IT JUST MEANS IT LOADS ON DEMAND IT ONLY LOADS WHEN IT NEEDS A '
51
+ # "PARTICULAR TYPE OF FILE LIKE WHEN IT'S BEING ACCESSED YEAH I THINK "
52
+ # "THAT'S THE IDEA IT JUST LOADS THE PARTICULAR ONES IT NEEDS BUT IF "
53
+ # "YOU WERE DOING A SEARCH OVER THE WHOLE CORPUS YOU'D HAVE TO LOAD "
54
+ # 'THEM ALL HMM',
55
+ # 'text_norm': 'has has anyone actually looked at the java code for the huh '
56
+ # 'yeah i think so yeah i i do not know about the search '
57
+ # 'functionality that might be online depends how it is going to '
58
+ # 'work yeah yeah that makes sense yeah you just concatenate them '
59
+ # 'together yeah it just means it loads on demand it only loads '
60
+ # 'when it needs a particular type of file like when it is being '
61
+ # 'accessed yeah i think that is the idea it just loads the '
62
+ # 'particular ones it needs but if you were doing a search over '
63
+ # 'the whole corpus you would have to load them all',
64
+ # 'wer': 4.716980934143066,
65
+ # 'whisper_transcript': '<|0.00|> Has anyone actually looked at the Java code '
66
+ # 'for the AMX?<|5.00|><|5.38|> Yeah, I think '
67
+ # "so.<|6.22|><|6.22|> Yeah, I don't know about the "
68
+ # 'search functionality.<|8.28|><|8.28|> That might be '
69
+ # "online.<|10.20|><|10.20|> Depends how it's gonna "
70
+ # 'work.<|11.92|><|11.92|> Yeah, that makes '
71
+ # 'sense.<|13.22|><|13.22|> Yeah, you just concatenate '
72
+ # 'them together.<|15.60|><|15.60|> It just means it '
73
+ # 'loads on demand.<|17.42|><|17.42|> It only loads when '
74
+ # 'it needs a particular type of file,<|22.24|><|22.24|> '
75
+ # "like when it's being accessed.<|23.40|><|23.40|> Yeah, "
76
+ # "I think that's the idea.<|24.40|><|24.40|> It just "
77
+ # 'loads the particular ones it needs.<|26.96|><|26.96|> '
78
+ # 'But if you were doing a search over the whole '
79
+ # "corpus,<|28.66|><|28.66|> you'd have to load them "
80
+ # 'all.<|29.96|>',
81
+ # 'whisper_transcript_norm': 'has anyone actually looked at the java code for '
82
+ # 'the amx yeah i think so yeah i do not know about '
83
+ # 'the search functionality that might be online '
84
+ # 'depends how it is going to work yeah that makes '
85
+ # 'sense yeah you just concatenate them together it '
86
+ # 'just means it loads on demand it only loads when '
87
+ # 'it needs a particular type of file like when it '
88
+ # 'is being accessed yeah i think that is the idea '
89
+ # 'it just loads the particular ones it needs but if '
90
+ # 'you were doing a search over the whole corpus you '
91
+ # 'would have to load them all'}
92
+ ```
93
+
94
+ ## Statistics
95
+
96
+ See below for the durations (in hours) after applying different WER filters to each subset.
97
+
98
+ English
99
+
100
+ | Split | 20% | 10% | 5% | 0% |
101
+ | :--- | :---: | :---: | :---: | :---: |
102
+ | en-mcv | 1,571.18 | 1,527.70 | 1,181.29 | 428.52 |
103
+ | en-ls | 951.31 | 932.31 | 852.46 | 450.89 |
104
+ | en-voxpopuli | 494.10 | 413.92 | 260.32 | 74.07 |
105
+ | en-tedlium | 448.05 | 416.95 | 312.36 | 78.16 |
106
+ | en-peoples_speech-clean | 5,652.32 | 3,474.44 | 1,160.46 | 73.02 |
107
+ | en-peoples_speech-clean_sa | 955.14 | 643.65 | 260.26 | 24.91 |
108
+ | en-peoples_speech-dirty | 8,664.09 | 1,414.41 | 181.96 | 5.35 |
109
+ | en-peoples_speech-dirty_sa | 972.02 | 206.29 | 35.34 | 1.65 |
110
+ | en-gigaspeech-l | 2,464.07 | 2,384.80 | 2,099.40 | 901.77 |
111
+ | en-ami-ihm | 50.84 | 19.54 | 5.44 | 0.46 |
112
+ | en-ami-sdm | 23.17 | 6.81 | 1.84 | 0.18 |
113
+ | en-yodas-000 | 3,699.62 | 2,902.96 | 1,891.37 | 487.83 |
114
+ | en-yodas-001 | 3,693.85 | 2,896.02 | 1,887.55 | 484.64 |
115
+ | en-yodas-002 | 3,687.30 | 2,890.38 | 1,895.27 | 487.38 |
116
+ | en-yodas-003 | 3,650.57 | 2,843.52 | 1,841.51 | 464.97 |
117
+ | en-yodas-004 | 3,710.20 | 2,907.12 | 1,890.28 | 477.85 |
118
+ | en-yodas-005 | 2,936.86 | 2,302.64 | 1,496.23 | 382.48 |
119
+ | en-yodas-100 | 3,831.71 | 2,692.86 | 1,496.27 | 286.69 |
120
+ | en-yodas-101 | 3,816.33 | 2,689.93 | 1,497.48 | 292.58 |
121
+ | en-yodas-102 | 3,826.86 | 2,701.17 | 1,501.30 | 286.49 |
122
+ | en-yodas-103 | 3,825.10 | 2,698.54 | 1,498.18 | 294.47 |
123
+ | en-yodas-104 | 2,449.36 | 1,717.38 | 948.36 | 184.66 |
124
+ | en-yodas-105 | 3,790.39 | 2,664.18 | 1,476.47 | 285.18 |
125
+ | en-yodas-106 | 3,800.00 | 2,678.34 | 1,487.09 | 287.32 |
126
+ | en-yodas-107 | 3,809.05 | 2,679.41 | 1,488.62 | 289.25 |
127
+ | en-yodas-109 | 3,791.46 | 2,677.90 | 1,492.26 | 290.38 |
128
+ | en-yodas-110 | 3,767.50 | 2,638.29 | 1,456.01 | 281.24 |
129
+ | en-yodas-111 | 3,801.11 | 2,671.89 | 1,486.27 | 287.74 |
130
+ | en-yodas-112 | 3,827.94 | 2,696.10 | 1,494.42 | 285.60 |
131
+ | en-yodas-113 | 3,817.43 | 2,681.09 | 1,489.73 | 289.74 |
132
+ | en-yodas-114 | 3,798.91 | 2,682.03 | 1,500.48 | 296.98 |
133
+ | en-yodas-115 | 3,811.49 | 2,682.46 | 1,487.23 | 288.86 |
134
+ | en-yodas-116 | 3,826.62 | 2,706.08 | 1,509.38 | 292.36 |
135
+ | en-yodas-117 | 3,808.30 | 2,684.16 | 1,497.32 | 293.84 |
136
+ | en-yodas-118 | 3,804.02 | 2,687.24 | 1,499.53 | 292.64 |
137
+ | en-yodas-119 | 3,809.40 | 2,697.34 | 1,508.54 | 296.01 |
138
+ | en-yodas-120 | 3,827.33 | 2,701.14 | 1,502.67 | 287.27 |
139
+ | en-yodas-121 | 3,800.26 | 2,677.95 | 1,488.84 | 290.16 |
140
+ | en-yodas-122 | 3,790.63 | 2,660.88 | 1,472.27 | 285.75 |
141
+ | en-yodas-123 | 3,785.27 | 2,677.23 | 1,494.34 | 289.47 |
142
+ | en-yodas-124 | 3,809.97 | 2,685.33 | 1,501.46 | 293.05 |
143
+ | en-yodas-125 | 3,783.51 | 2,659.93 | 1,475.39 | 288.66 |
144
+ | en-yodas-126 | 3,797.07 | 2,668.46 | 1,487.35 | 289.60 |
145
+ | en-yodas-127 | 1,769.64 | 1,247.31 | 699.89 | 137.38 |
146
+ | total | 143,001.36 | 98,188.11 | 56,190.49 | 12,387.48 |
147
+
148
+ French
149
+
150
+ | Split | 20% | 10% | 5% | 0% |
151
+ | :--- | :---: | :---: | :---: | :---: |
152
+ | fr-mcv | 689.80 | 663.32 | 439.61 | 93.34 |
153
+ | fr-mls | 1,042.59 | 936.38 | 703.29 | 260.22 |
154
+ | fr-voxpopuli | 191.70 | 146.51 | 84.15 | 21.91 |
155
+ | fr-mtedx | 146.09 | 100.67 | 57.22 | 12.98 |
156
+ | fr-yodas-000 | 1,497.83 | 912.55 | 445.32 | 71.56 |
157
+ | fr-yodas-100 | 1,860.75 | 606.79 | 149.23 | 13.01 |
158
+ | fr-yodas-101 | 1,857.40 | 612.54 | 151.96 | 14.09 |
159
+ | fr-yodas-102 | 1,850.93 | 610.35 | 152.34 | 13.33 |
160
+ | fr-yodas-103 | 1,172.29 | 390.89 | 98.22 | 9.21 |
161
+ | total | 10,309.39 | 4,979.99 | 2,281.33 | 509.65 |
162
+
163
+ Spanish
164
+
165
+ | Split | 20% | 10% | 5% | 0% |
166
+ | :--- | :---: | :---: | :---: | :---: |
167
+ | es-mcv | 446.01 | 435.19 | 350.02 | 145.81 |
168
+ | es-mls | 844.79 | 722.04 | 535.54 | 210.70 |
169
+ | es-voxpopuli | 139.52 | 112.18 | 70.99 | 20.37 |
170
+ | es-mtedx | 150.89 | 114.39 | 68.60 | 16.38 |
171
+ | es-yodas-000 | 2,408.25 | 1,592.19 | 851.84 | 180.98 |
172
+ | es-yodas-100 | 2,982.87 | 1,610.76 | 667.60 | 104.08 |
173
+ | es-yodas-101 | 2,987.42 | 1,584.88 | 647.20 | 100.99 |
174
+ | total | 9,959.76 | 6,171.63 | 3,191.77 | 779.30 |
175
+
176
+ Portuguese
177
+
178
+ | Split | 20% | 10% | 5% | 0% |
179
+ | :--- | :---: | :---: | :---: | :---: |
180
+ | pt-mcv | 21.75 | 21.42 | 19.13 | 10.46 |
181
+ | pt-mls | 147.01 | 113.41 | 69.66 | 20.76 |
182
+ | pt-mtedx | 131.71 | 94.51 | 49.57 | 9.32 |
183
+ | pt-yodas-000 | 859.83 | 453.14 | 211.42 | 42.00 |
184
+ | pt-yodas-100 | 1,853.79 | 549.13 | 140.03 | 22.36 |
185
+ | pt-yodas-101 | 1,849.31 | 552.92 | 141.34 | 21.55 |
186
+ | pt-yodas-102 | 1,871.89 | 560.40 | 143.63 | 22.47 |
187
+ | pt-yodas-103 | 1,288.90 | 383.86 | 98.71 | 15.83 |
188
+ | total | 8,024.19 | 2,728.80 | 873.48 | 164.75 |
189
+
190
+ Italian
191
+
192
+ | Split | 20% | 10% | 5% | 0% |
193
+ | :--- | :---: | :---: | :---: | :---: |
194
+ | it-mcv | 232.83 | 229.54 | 187.96 | 70.52 |
195
+ | it-mls | 232.95 | 185.26 | 113.87 | 35.30 |
196
+ | it-voxpopuli | 58.23 | 41.64 | 23.06 | 6.31 |
197
+ | it-mtedx | 88.72 | 73.84 | 49.47 | 13.02 |
198
+ | it-yodas-000 | 952.76 | 600.62 | 317.58 | 85.28 |
199
+ | it-yodas-100 | 2,664.50 | 1,242.66 | 453.31 | 70.46 |
200
+ | it-yodas-101 | 2,277.28 | 1,062.83 | 387.99 | 60.38 |
201
+ | total | 6,507.27 | 3,436.38 | 1,533.24 | 341.27 |
202
+
203
+ German
204
+
205
+ | Split | 20% | 10% | 5% | 0% |
206
+ | :--- | :---: | :---: | :---: | :---: |
207
+ | de-mcv | 875.27 | 862.86 | 720.29 | 324.03 |
208
+ | de-mls | 1,919.13 | 1,736.71 | 1,315.35 | 661.96 |
209
+ | de-voxpopuli | 232.27 | 146.60 | 70.01 | 17.32 |
210
+ | de-mtedx | 8.39 | 5.70 | 3.28 | 0.81 |
211
+ | de-yodas-000 | 1,607.82 | 925.94 | 476.64 | 128.54 |
212
+ | de-yodas-100 | 2,304.63 | 856.85 | 260.56 | 41.08 |
213
+ | de-yodas-101 | 2,343.04 | 875.50 | 265.74 | 40.33 |
214
+ | de-yodas-102 | 426.58 | 156.51 | 47.43 | 7.58 |
215
+ | total | 9,717.13 | 5,566.67 | 3,159.30 | 1,221.65 |
216
+
217
+ Dutch
218
+
219
+ | Split | 20% | 10% | 5% | 0% |
220
+ | :--- | :---: | :---: | :---: | :---: |
221
+ | nl-mcv | 41.02 | 40.81 | 35.81 | 12.01 |
222
+ | nl-mls | 1,455.24 | 1,133.02 | 692.95 | 253.02 |
223
+ | nl-voxpopuli | 38.86 | 23.20 | 9.50 | 1.79 |
224
+ | nl-yodas-000 | 215.26 | 108.59 | 44.07 | 6.59 |
225
+ | nl-yodas-100 | 512.37 | 127.16 | 31.75 | 5.48 |
226
+ | total | 2,262.75 | 1,432.78 | 814.08 | 278.89 |
227
+
228
+ Code-switching
229
+
230
+ | Split | 20% | 10% | 5% | 0% |
231
+ | :--- | :---: | :---: | :---: | :---: |
232
+ | cs-mcv | 3,346.80 | 3,333.61 | 3,192.75 | 1,428.04 |
233
+ | cs-yodas | 6,465.59 | 6,434.34 | 4,967.96 | 572.10 |
234
+ | total | 9,812.39 | 9,767.94 | 8,160.72 | 2,000.14 |
__pycache__/meta.cpython-310.pyc ADDED
Binary file (279 kB). View file
 
dur.txt ADDED
@@ -0,0 +1,138 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ it
3
+
4
+ | Split | 20% | 10% | 5% | 0% |
5
+ | :--- | :---: | :---: | :---: | :---: |
6
+ | it-mcv | 232.83 | 229.54 | 187.96 | 70.52 |
7
+ | it-mls | 232.95 | 185.26 | 113.87 | 35.30 |
8
+ | it-voxpopuli | 58.23 | 41.64 | 23.06 | 6.31 |
9
+ | it-mtedx | 88.72 | 73.84 | 49.47 | 13.02 |
10
+ | it-yodas-000 | 952.76 | 600.62 | 317.58 | 85.28 |
11
+ | it-yodas-100 | 2,664.50 | 1,242.66 | 453.31 | 70.46 |
12
+ | it-yodas-101 | 2,277.28 | 1,062.83 | 387.99 | 60.38 |
13
+ | total | 6,507.27 | 3,436.38 | 1,533.24 | 341.27 |
14
+
15
+ fr
16
+
17
+ | Split | 20% | 10% | 5% | 0% |
18
+ | :--- | :---: | :---: | :---: | :---: |
19
+ | fr-mcv | 689.80 | 663.32 | 439.61 | 93.34 |
20
+ | fr-mls | 1,042.59 | 936.38 | 703.29 | 260.22 |
21
+ | fr-voxpopuli | 191.70 | 146.51 | 84.15 | 21.91 |
22
+ | fr-mtedx | 146.09 | 100.67 | 57.22 | 12.98 |
23
+ | fr-yodas-000 | 1,497.83 | 912.55 | 445.32 | 71.56 |
24
+ | fr-yodas-100 | 1,860.75 | 606.79 | 149.23 | 13.01 |
25
+ | fr-yodas-101 | 1,857.40 | 612.54 | 151.96 | 14.09 |
26
+ | fr-yodas-102 | 1,850.93 | 610.35 | 152.34 | 13.33 |
27
+ | fr-yodas-103 | 1,172.29 | 390.89 | 98.22 | 9.21 |
28
+ | total | 10,309.39 | 4,979.99 | 2,281.33 | 509.65 |
29
+
30
+ es
31
+
32
+ | Split | 20% | 10% | 5% | 0% |
33
+ | :--- | :---: | :---: | :---: | :---: |
34
+ | es-mcv | 446.01 | 435.19 | 350.02 | 145.81 |
35
+ | es-mls | 844.79 | 722.04 | 535.54 | 210.70 |
36
+ | es-voxpopuli | 139.52 | 112.18 | 70.99 | 20.37 |
37
+ | es-mtedx | 150.89 | 114.39 | 68.60 | 16.38 |
38
+ | es-yodas-000 | 2,408.25 | 1,592.19 | 851.84 | 180.98 |
39
+ | es-yodas-100 | 2,982.87 | 1,610.76 | 667.60 | 104.08 |
40
+ | es-yodas-101 | 2,987.42 | 1,584.88 | 647.20 | 100.99 |
41
+ | total | 9,959.76 | 6,171.63 | 3,191.77 | 779.30 |
42
+
43
+ pt
44
+
45
+ | Split | 20% | 10% | 5% | 0% |
46
+ | :--- | :---: | :---: | :---: | :---: |
47
+ | pt-mcv | 21.75 | 21.42 | 19.13 | 10.46 |
48
+ | pt-mls | 147.01 | 113.41 | 69.66 | 20.76 |
49
+ | pt-mtedx | 131.71 | 94.51 | 49.57 | 9.32 |
50
+ | pt-yodas-000 | 859.83 | 453.14 | 211.42 | 42.00 |
51
+ | pt-yodas-100 | 1,853.79 | 549.13 | 140.03 | 22.36 |
52
+ | pt-yodas-101 | 1,849.31 | 552.92 | 141.34 | 21.55 |
53
+ | pt-yodas-102 | 1,871.89 | 560.40 | 143.63 | 22.47 |
54
+ | pt-yodas-103 | 1,288.90 | 383.86 | 98.71 | 15.83 |
55
+ | total | 8,024.19 | 2,728.80 | 873.48 | 164.75 |
56
+
57
+ de
58
+
59
+ | Split | 20% | 10% | 5% | 0% |
60
+ | :--- | :---: | :---: | :---: | :---: |
61
+ | de-mcv | 875.27 | 862.86 | 720.29 | 324.03 |
62
+ | de-mls | 1,919.13 | 1,736.71 | 1,315.35 | 661.96 |
63
+ | de-voxpopuli | 232.27 | 146.60 | 70.01 | 17.32 |
64
+ | de-mtedx | 8.39 | 5.70 | 3.28 | 0.81 |
65
+ | de-yodas-000 | 1,607.82 | 925.94 | 476.64 | 128.54 |
66
+ | de-yodas-100 | 2,304.63 | 856.85 | 260.56 | 41.08 |
67
+ | de-yodas-101 | 2,343.04 | 875.50 | 265.74 | 40.33 |
68
+ | de-yodas-102 | 426.58 | 156.51 | 47.43 | 7.58 |
69
+ | total | 9,717.13 | 5,566.67 | 3,159.30 | 1,221.65 |
70
+
71
+ nl
72
+
73
+ | Split | 20% | 10% | 5% | 0% |
74
+ | :--- | :---: | :---: | :---: | :---: |
75
+ | nl-mcv | 41.02 | 40.81 | 35.81 | 12.01 |
76
+ | nl-mls | 1,455.24 | 1,133.02 | 692.95 | 253.02 |
77
+ | nl-voxpopuli | 38.86 | 23.20 | 9.50 | 1.79 |
78
+ | nl-yodas-000 | 215.26 | 108.59 | 44.07 | 6.59 |
79
+ | nl-yodas-100 | 512.37 | 127.16 | 31.75 | 5.48 |
80
+ | total | 2,262.75 | 1,432.78 | 814.08 | 278.89 |
81
+
82
+ en
83
+
84
+ | Split | 20% | 10% | 5% | 0% |
85
+ | :--- | :---: | :---: | :---: | :---: |
86
+ | en-mcv | 1,571.18 | 1,527.70 | 1,181.29 | 428.52 |
87
+ | en-ls | 951.31 | 932.31 | 852.46 | 450.89 |
88
+ | en-voxpopuli | 494.10 | 413.92 | 260.32 | 74.07 |
89
+ | en-tedlium | 448.05 | 416.95 | 312.36 | 78.16 |
90
+ | en-peoples_speech-clean | 5,652.32 | 3,474.44 | 1,160.46 | 73.02 |
91
+ | en-peoples_speech-clean_sa | 955.14 | 643.65 | 260.26 | 24.91 |
92
+ | en-peoples_speech-dirty | 8,664.09 | 1,414.41 | 181.96 | 5.35 |
93
+ | en-peoples_speech-dirty_sa | 972.02 | 206.29 | 35.34 | 1.65 |
94
+ | en-gigaspeech-l | 2,464.07 | 2,384.80 | 2,099.40 | 901.77 |
95
+ | en-ami-ihm | 50.84 | 19.54 | 5.44 | 0.46 |
96
+ | en-ami-sdm | 23.17 | 6.81 | 1.84 | 0.18 |
97
+ | en-yodas-000 | 3,699.62 | 2,902.96 | 1,891.37 | 487.83 |
98
+ | en-yodas-001 | 3,693.85 | 2,896.02 | 1,887.55 | 484.64 |
99
+ | en-yodas-002 | 3,687.30 | 2,890.38 | 1,895.27 | 487.38 |
100
+ | en-yodas-003 | 3,650.57 | 2,843.52 | 1,841.51 | 464.97 |
101
+ | en-yodas-004 | 3,710.20 | 2,907.12 | 1,890.28 | 477.85 |
102
+ | en-yodas-005 | 2,936.86 | 2,302.64 | 1,496.23 | 382.48 |
103
+ | en-yodas-100 | 3,831.71 | 2,692.86 | 1,496.27 | 286.69 |
104
+ | en-yodas-101 | 3,816.33 | 2,689.93 | 1,497.48 | 292.58 |
105
+ | en-yodas-102 | 3,826.86 | 2,701.17 | 1,501.30 | 286.49 |
106
+ | en-yodas-103 | 3,825.10 | 2,698.54 | 1,498.18 | 294.47 |
107
+ | en-yodas-104 | 2,449.36 | 1,717.38 | 948.36 | 184.66 |
108
+ | en-yodas-105 | 3,790.39 | 2,664.18 | 1,476.47 | 285.18 |
109
+ | en-yodas-106 | 3,800.00 | 2,678.34 | 1,487.09 | 287.32 |
110
+ | en-yodas-107 | 3,809.05 | 2,679.41 | 1,488.62 | 289.25 |
111
+ | en-yodas-109 | 3,791.46 | 2,677.90 | 1,492.26 | 290.38 |
112
+ | en-yodas-110 | 3,767.50 | 2,638.29 | 1,456.01 | 281.24 |
113
+ | en-yodas-111 | 3,801.11 | 2,671.89 | 1,486.27 | 287.74 |
114
+ | en-yodas-112 | 3,827.94 | 2,696.10 | 1,494.42 | 285.60 |
115
+ | en-yodas-113 | 3,817.43 | 2,681.09 | 1,489.73 | 289.74 |
116
+ | en-yodas-114 | 3,798.91 | 2,682.03 | 1,500.48 | 296.98 |
117
+ | en-yodas-115 | 3,811.49 | 2,682.46 | 1,487.23 | 288.86 |
118
+ | en-yodas-116 | 3,826.62 | 2,706.08 | 1,509.38 | 292.36 |
119
+ | en-yodas-117 | 3,808.30 | 2,684.16 | 1,497.32 | 293.84 |
120
+ | en-yodas-118 | 3,804.02 | 2,687.24 | 1,499.53 | 292.64 |
121
+ | en-yodas-119 | 3,809.40 | 2,697.34 | 1,508.54 | 296.01 |
122
+ | en-yodas-120 | 3,827.33 | 2,701.14 | 1,502.67 | 287.27 |
123
+ | en-yodas-121 | 3,800.26 | 2,677.95 | 1,488.84 | 290.16 |
124
+ | en-yodas-122 | 3,790.63 | 2,660.88 | 1,472.27 | 285.75 |
125
+ | en-yodas-123 | 3,785.27 | 2,677.23 | 1,494.34 | 289.47 |
126
+ | en-yodas-124 | 3,809.97 | 2,685.33 | 1,501.46 | 293.05 |
127
+ | en-yodas-125 | 3,783.51 | 2,659.93 | 1,475.39 | 288.66 |
128
+ | en-yodas-126 | 3,797.07 | 2,668.46 | 1,487.35 | 289.60 |
129
+ | en-yodas-127 | 1,769.64 | 1,247.31 | 699.89 | 137.38 |
130
+ | total | 143,001.36 | 98,188.11 | 56,190.49 | 12,387.48 |
131
+
132
+ cs
133
+
134
+ | Split | 20% | 10% | 5% | 0% |
135
+ | :--- | :---: | :---: | :---: | :---: |
136
+ | cs-mcv | 3,346.80 | 3,333.61 | 3,192.75 | 1,428.04 |
137
+ | cs-yodas | 6,465.59 | 6,434.34 | 4,967.96 | 572.10 |
138
+ | total | 9,812.39 | 9,767.94 | 8,160.72 | 2,000.14 |
get_stats.py ADDED
@@ -0,0 +1,84 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python
2
+ # coding=utf-8
3
+ # Copyright 2024 Bofeng Huang
4
+
5
+ import os
6
+ from collections import defaultdict
7
+
8
+ import fire
9
+ import numpy as np
10
+ import pandas as pd
11
+ from tqdm import tqdm
12
+
13
+ from meta import SUBSET_NAMES_AND_PATHS
14
+
15
+
16
+ def _print_ds_info(df, duration_column_name="duration"):
17
+ print(f"#utterances: {df.shape[0]}")
18
+ durations = df["duration"]
19
+ print(
20
+ f"Duration statistics: tot {durations.sum() / 3600:.2f} h, "
21
+ f"mean {durations.mean():.2f} s, "
22
+ f"min {durations.min():.2f} s, "
23
+ f"max {durations.max():.2f} s"
24
+ )
25
+ print()
26
+
27
+
28
+ def main(output_file):
29
+
30
+ dataset_dir = os.path.dirname(os.path.abspath(__file__))
31
+
32
+ lang_manifests_dict = defaultdict(list)
33
+ for k, v in SUBSET_NAMES_AND_PATHS.items():
34
+ lang_manifests_dict[k.split("-")[0]].append((k, f'{dataset_dir}/{v["dir"]}/{v["text_file"]}'))
35
+
36
+ # print(lang_manifests_dict)
37
+
38
+ with open(output_file, "w") as f:
39
+ for lang, manifest_files in lang_manifests_dict.items():
40
+ f.write("\n" + lang + "\n" + "\n")
41
+ f.write("| Split | 20% | 10% | 5% | 0% |" + "\n")
42
+ f.write("| :--- | :---: | :---: | :---: | :---: |" + "\n")
43
+
44
+ lines = []
45
+ for split, manifest_file in tqdm(manifest_files):
46
+ # load dataset
47
+ df = pd.read_json(manifest_file, lines=True)
48
+ # print("Raw dataset")
49
+ # _print_ds_info(df)
50
+
51
+ # line = f"| {split} |"
52
+
53
+ # wer_cutoffs = [20, 10, 5, 0]
54
+ # for wer_cutoff in wer_cutoffs:
55
+ # df_ = df[df["wer"] <= wer_cutoff]
56
+ # # print(f"wer_cutoff: {wer_cutoff}")
57
+ # # _print_ds_info(df_)
58
+
59
+ # line += f' {df_["duration"].sum() / 3600:.2f} |'
60
+
61
+ # f.write(line + "\n")
62
+
63
+ l = [df[df["wer"] <= wer_cutoff]["duration"].sum() / 3600 for wer_cutoff in [20, 10, 5, 0]]
64
+ l.insert(0, split)
65
+ lines.append(l)
66
+
67
+ lines.append(
68
+ [
69
+ "total",
70
+ sum(l[1] for l in lines),
71
+ sum(l[2] for l in lines),
72
+ sum(l[3] for l in lines),
73
+ sum(l[4] for l in lines),
74
+ ]
75
+ )
76
+
77
+ for l in lines:
78
+ f.write(f"| {l[0]} | " + " | ".join([f"{l_:,.2f}" for l_ in l[1:]]) + " |" + "\n")
79
+
80
+ # break
81
+
82
+
83
+ if __name__ == "__main__":
84
+ fire.Fire(main)