tobiolatunji commited on
Commit
a239d5a
1 Parent(s): 41181f7

upload tar files and csvs

Browse files
Files changed (46) hide show
  1. .gitattributes +1 -0
  2. README.md +364 -0
  3. afrispeech-200.py +160 -0
  4. audio/dev/dev_0.tar.gz +3 -0
  5. audio/dev/dev_1.tar.gz +3 -0
  6. audio/test/test_0.tar.gz +3 -0
  7. audio/test/test_1.tar.gz +3 -0
  8. audio/test/test_2.tar.gz +3 -0
  9. audio/test/test_3.tar.gz +3 -0
  10. audio/train/train_0.tar.gz +3 -0
  11. audio/train/train_1.tar.gz +3 -0
  12. audio/train/train_10.tar.gz +3 -0
  13. audio/train/train_11.tar.gz +3 -0
  14. audio/train/train_12.tar.gz +3 -0
  15. audio/train/train_13.tar.gz +3 -0
  16. audio/train/train_14.tar.gz +3 -0
  17. audio/train/train_15.tar.gz +3 -0
  18. audio/train/train_16.tar.gz +3 -0
  19. audio/train/train_17.tar.gz +3 -0
  20. audio/train/train_18.tar.gz +3 -0
  21. audio/train/train_19.tar.gz +3 -0
  22. audio/train/train_2.tar.gz +3 -0
  23. audio/train/train_20.tar.gz +3 -0
  24. audio/train/train_21.tar.gz +3 -0
  25. audio/train/train_22.tar.gz +3 -0
  26. audio/train/train_23.tar.gz +3 -0
  27. audio/train/train_24.tar.gz +3 -0
  28. audio/train/train_25.tar.gz +3 -0
  29. audio/train/train_26.tar.gz +3 -0
  30. audio/train/train_27.tar.gz +3 -0
  31. audio/train/train_28.tar.gz +3 -0
  32. audio/train/train_29.tar.gz +3 -0
  33. audio/train/train_3.tar.gz +3 -0
  34. audio/train/train_30.tar.gz +3 -0
  35. audio/train/train_31.tar.gz +3 -0
  36. audio/train/train_32.tar.gz +3 -0
  37. audio/train/train_33.tar.gz +3 -0
  38. audio/train/train_34.tar.gz +3 -0
  39. audio/train/train_4.tar.gz +3 -0
  40. audio/train/train_5.tar.gz +3 -0
  41. audio/train/train_6.tar.gz +3 -0
  42. audio/train/train_7.tar.gz +3 -0
  43. audio/train/train_8.tar.gz +3 -0
  44. audio/train/train_9.tar.gz +3 -0
  45. transcripts/dev.csv +3 -0
  46. transcripts/test.csv +3 -0
.gitattributes CHANGED
@@ -52,3 +52,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
52
  *.jpg filter=lfs diff=lfs merge=lfs -text
53
  *.jpeg filter=lfs diff=lfs merge=lfs -text
54
  *.webp filter=lfs diff=lfs merge=lfs -text
 
 
52
  *.jpg filter=lfs diff=lfs merge=lfs -text
53
  *.jpeg filter=lfs diff=lfs merge=lfs -text
54
  *.webp filter=lfs diff=lfs merge=lfs -text
55
+ *.csv filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,364 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ pretty_name: AfriSpeech-200
3
+ annotations_creators:
4
+ - expert-generated
5
+ language_creators:
6
+ - crowdsourced
7
+ - expert-generated
8
+ language:
9
+ - en
10
+ license:
11
+ - cc-by-nc-sa-4.0
12
+ multilinguality:
13
+ - monolingual
14
+ size_categories:
15
+ - 10K<n<100K
16
+ source_datasets:
17
+ - original
18
+ task_categories:
19
+ - automatic-speech-recognition
20
+ task_ids: []
21
+ dataset_info:
22
+ features:
23
+ - name: user_id
24
+ dtype: string
25
+ - name: path
26
+ dtype: string
27
+ - name: audio
28
+ dtype:
29
+ audio:
30
+ sampling_rate: 44100
31
+ - name: transcript
32
+ dtype: string
33
+ splits:
34
+ - name: train
35
+ num_bytes: 1722002133
36
+ num_examples: 58000
37
+ - name: dev
38
+ num_bytes: 86120227
39
+ num_examples: 3231
40
+ download_size: 1475540500
41
+ dataset_size: 1808122360
42
+ extra_gated_prompt: By clicking on “Access repository” below, you also agree to not attempt to determine the
43
+ identity of speakers in the Common Voice dataset.
44
+ ---
45
+
46
+ # Dataset Card for AfriSpeech-200
47
+
48
+ ## Table of Contents
49
+ - [Dataset Card for AfriSpeech-200](#dataset-card-for-afrispeech-200)
50
+ - [Table of Contents](#table-of-contents)
51
+ - [Dataset Description](#dataset-description)
52
+ - [Dataset Summary](#dataset-summary)
53
+ - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
54
+ - [Languages](#languages)
55
+ - [Dataset Structure](#dataset-structure)
56
+ - [Data Instances](#data-instances)
57
+ - [Data Fields](#data-fields)
58
+ - [Data Splits](#data-splits)
59
+ - [Dataset Creation](#dataset-creation)
60
+ - [Curation Rationale](#curation-rationale)
61
+ - [Source Data](#source-data)
62
+ - [Initial Data Collection and Normalization](#initial-data-collection-and-normalization)
63
+ - [Who are the source language producers?](#who-are-the-source-language-producers)
64
+ - [Annotations](#annotations)
65
+ - [Annotation process](#annotation-process)
66
+ - [Who are the annotators?](#who-are-the-annotators)
67
+ - [Personal and Sensitive Information](#personal-and-sensitive-information)
68
+ - [Considerations for Using the Data](#considerations-for-using-the-data)
69
+ - [Social Impact of Dataset](#social-impact-of-dataset)
70
+ - [Discussion of Biases](#discussion-of-biases)
71
+ - [Other Known Limitations](#other-known-limitations)
72
+ - [Additional Information](#additional-information)
73
+ - [Dataset Curators](#dataset-curators)
74
+ - [Licensing Information](#licensing-information)
75
+ - [Citation Information](#citation-information)
76
+ - [Contributions](#contributions)
77
+
78
+ ## Dataset Description
79
+
80
+ - **Homepage:** https://github.com/intron-innovation/AfriSpeech-Dataset-Paper
81
+ - **Repository:** https://github.com/intron-innovation/AfriSpeech-Dataset-Paper
82
+ - **Paper:** [AfriSpeech-200: Pan-African accented speech dataset for clinical and general domain ASR](https://github.com/intron-innovation/AfriSpeech-Dataset-Paper)
83
+ - **Leaderboard:** [Needs More Information]
84
+ - **Point of Contact:** [Intron Innovation](mailto:intron@intron.io)
85
+
86
+ ### Dataset Summary
87
+
88
+ AFRISPEECH-200 is a 200hr Pan-African speech corpus for clinical and general domain English accented ASR; a dataset with 120 African accents from 13 countries and 2,463 unique African speakers.
89
+ Our goal is to raise awareness for and advance Pan-African English ASR research, especially for the clinical domain.
90
+
91
+ ### Supported Tasks and Leaderboards
92
+
93
+ - Automatic Speech Recognition
94
+ - Speech Synthesis (Text-to-Speech)
95
+
96
+ ### Languages
97
+
98
+ English (Accented)
99
+
100
+ ## Dataset Structure
101
+
102
+ ### Data Instances
103
+
104
+ A typical data point comprises the path to the audio file, called `path` and its transcription, called `sentence`. Some additional information about the speaker and the passage which contains the transcription is provided.
105
+
106
+ ```
107
+ {'user_id': 'ABCDXYZ01',
108
+ 'path': '/home/admin/.cache/huggingface/datasets/downloads/extracted/b7ded9969e09942ab65313e691e6fc2e12066192ee8527e21d634aca128afbe2/vivos/train/waves/VIVOSSPK01/VIVOSSPK01_R001.wav',
109
+ 'audio': {'path': '/home/admin/.cache/huggingface/datasets/downloads/extracted/b7ded9969e09942ab65313e691e6fc2e12066192ee8527e21d634aca128afbe2/vivos/train/waves/VIVOSSPK01/VIVOSSPK01_R001.wav',
110
+ 'array': array([-0.00048828, -0.00018311, -0.00137329, ..., 0.00079346, 0.00091553, 0.00085449], dtype=float32),
111
+ 'sampling_rate': 16000},
112
+ 'transcirpt': 'The patient took the correct medication'}
113
+ ```
114
+
115
+ ### Data Fields
116
+
117
+ - speaker_id: An id for which speaker (voice) made the recording
118
+
119
+ - path: The path to the audio file
120
+
121
+ - audio: A dictionary containing the path to the downloaded audio file, the decoded audio array, and the sampling rate. Note that when accessing the audio column: `dataset[0]["audio"]` the audio file is automatically decoded and resampled to `dataset.features["audio"].sampling_rate`. Decoding and resampling of a large number of audio files might take a significant amount of time. Thus it is important to first query the sample index before the `"audio"` column, *i.e.* `dataset[0]["audio"]` should **always** be preferred over `dataset["audio"][0]`.
122
+
123
+ - transcript: The sentence the user was prompted to speak
124
+
125
+ ### Data Splits
126
+
127
+ The speech material has been subdivided into portions for train, dev, and test.
128
+
129
+ Speech was recorded in a quiet environment with high quality microphone, speakers were asked to read one sentence at a time.
130
+
131
+ - Total Number of Unique Speakers: 2,463
132
+ - Female/Male/Other Ratio: 57.11/42.41/0.48
133
+ - Data was first split on speakers. Speakers in Train/Dev/Test do not cross partitions
134
+
135
+ | | Train | Dev | Test |
136
+ | ----------- | ----------- | ----------- | ----------- |
137
+ | # Speakers | 1466 | 247 | 750 |
138
+ | # Seconds | 624228.83 | 31447.09 | 67559.10 |
139
+ | # Hours | 173.4 | 8.74 | 18.77 |
140
+ | # Accents | 71 | 45 | 108 |
141
+ | Avg secs/speaker | 425.81 | 127.32 | 90.08 |
142
+ | Avg num clips/speaker | 39.56 | 13.08 | 8.46 |
143
+ | Avg num speakers/accent | 20.65 | 5.49 | 6.94 |
144
+ | Avg secs/accent | 8791.96 | 698.82 | 625.55 |
145
+ | # clips general domain | 21682 | 1407 | 2723 |
146
+ | # clips clinical domain | 36318 | 1824 | 3623 |
147
+
148
+ ## Dataset Creation
149
+
150
+ ### Curation Rationale
151
+
152
+ Africa has a very low doctor-to-patient ratio.
153
+ At very busy clinics, doctors could see 30+ patients per day-- a heavy patient burden compared with
154
+ developed countries-- but productivity tools such as clinical automatic speech recognition
155
+ (ASR) are lacking for these overworked clinicians. However, clinical ASR is mature, even ubiquitous,
156
+ in developed nations, and clinician-reported performance of commercial clinical ASR systems
157
+ is generally satisfactory. Furthermore, the recent performance of general domain ASR is
158
+ approaching human accuracy. However, several gaps exist. Several publications have
159
+ highlighted racial bias with speech-to-text algorithms and performance on minority
160
+ accents lags significantly. To our knowledge, there is no publicly available research or
161
+ benchmark on accented African clinical ASR, and speech data is non-existent for the
162
+ majority of African accents. We release AfriSpeech, 200hrs of Pan-African speech,
163
+ 67,577 clips from 2,463 unique speakers, across 120 indigenous accents from 13 countries for
164
+ clinical and general domain ASR, a benchmark test set, with publicly available pre-trained
165
+ models with SOTA performance on the AfriSpeech benchmark.
166
+
167
+ ### Source Data
168
+
169
+ #### Country Stats
170
+
171
+ | Country | Clips | Speakers | Duration (seconds) | Duration (hrs) |
172
+ | ----------- | ----------- | ----------- | ----------- | ----------- |
173
+ | NG | 45875 | 1979 | 512646.88 | 142.40 |
174
+ | KE | 8304 | 137 | 75195.43 | 20.89 |
175
+ | ZA | 7870 | 223 | 81688.11 | 22.69 |
176
+ | GH | 2018 | 37 | 18581.13 | 5.16 |
177
+ | BW | 1391 | 38 | 14249.01 | 3.96 |
178
+ | UG | 1092 | 26 | 10420.42 | 2.89 |
179
+ | RW | 469 | 9 | 5300.99 | 1.47 |
180
+ | US | 219 | 5 | 1900.98 | 0.53 |
181
+ | TR | 66 | 1 | 664.01 | 0.18 |
182
+ | ZW | 63 | 3 | 635.11 | 0.18 |
183
+ | MW | 60 | 1 | 554.61 | 0.15 |
184
+ | TZ | 51 | 2 | 645.51 | 0.18 |
185
+ | LS | 7 | 1 | 78.40 | 0.02 |
186
+
187
+ #### Accent Stats
188
+
189
+ | Accent | Clips | Speakers | Duration (s) | Country | Splits |
190
+ | ----------- | ----------- | ----------- | ----------- | ----------- | ----------- |
191
+ | yoruba | 15407 | 683 | 161587.55 | US,NG | train,test,dev |
192
+ | igbo | 8677 | 374 | 93035.79 | US,NG,ZA | train,test,dev |
193
+ | swahili | 6320 | 119 | 55932.82 | KE,TZ,ZA,UG | train,test,dev |
194
+ | hausa | 5765 | 248 | 70878.67 | NG | train,test,dev |
195
+ | ijaw | 2499 | 105 | 33178.9 | NG | train,test,dev |
196
+ | afrikaans | 2048 | 33 | 20586.49 | ZA | train,test,dev |
197
+ | idoma | 1877 | 72 | 20463.6 | NG | train,test,dev |
198
+ | zulu | 1794 | 52 | 18216.97 | ZA,TR,LS | dev,train,test |
199
+ | setswana | 1588 | 39 | 16553.22 | BW,ZA | dev,test,train |
200
+ | twi | 1566 | 22 | 14340.12 | GH | test,train,dev |
201
+ | isizulu | 1048 | 48 | 10376.09 | ZA | test,train,dev |
202
+ | igala | 919 | 31 | 9854.72 | NG | train,test |
203
+ | izon | 838 | 47 | 9602.53 | NG | train,dev,test |
204
+ | kiswahili | 827 | 6 | 8988.26 | KE | train,test |
205
+ | ebira | 757 | 42 | 7752.94 | NG | train,test,dev |
206
+ | luganda | 722 | 22 | 6768.19 | UG,BW,KE | test,dev,train |
207
+ | urhobo | 646 | 32 | 6685.12 | NG | train,dev,test |
208
+ | nembe | 578 | 16 | 6644.72 | NG | train,test,dev |
209
+ | ibibio | 570 | 39 | 6489.29 | NG | train,test,dev |
210
+ | pidgin | 514 | 20 | 5871.57 | NG | test,train,dev |
211
+ | luhya | 508 | 4 | 4497.02 | KE | train,test |
212
+ | kinyarwanda | 469 | 9 | 5300.99 | RW | train,test,dev |
213
+ | xhosa | 392 | 12 | 4604.84 | ZA | train,dev,test |
214
+ | tswana | 387 | 18 | 4148.58 | ZA,BW | train,test,dev |
215
+ | esan | 380 | 13 | 4162.63 | NG | train,test,dev |
216
+ | alago | 363 | 8 | 3902.09 | NG | train,test |
217
+ | tshivenda | 353 | 5 | 3264.77 | ZA | test,train |
218
+ | fulani | 312 | 18 | 5084.32 | NG | test,train |
219
+ | isoko | 298 | 16 | 4236.88 | NG | train,test,dev |
220
+ | akan (fante) | 295 | 9 | 2848.54 | GH | train,dev,test |
221
+ | ikwere | 293 | 14 | 3480.43 | NG | test,train,dev |
222
+ | sepedi | 275 | 10 | 2751.68 | ZA | dev,test,train |
223
+ | efik | 269 | 11 | 2559.32 | NG | test,train,dev |
224
+ | edo | 237 | 12 | 1842.32 | NG | train,test,dev |
225
+ | luo | 234 | 4 | 2052.25 | UG,KE | test,train,dev |
226
+ | kikuyu | 229 | 4 | 1949.62 | KE | train,test,dev |
227
+ | bekwarra | 218 | 3 | 2000.46 | NG | train,test |
228
+ | isixhosa | 210 | 9 | 2100.28 | ZA | train,dev,test |
229
+ | hausa/fulani | 202 | 3 | 2213.53 | NG | test,train |
230
+ | epie | 202 | 6 | 2320.21 | NG | train,test |
231
+ | isindebele | 198 | 2 | 1759.49 | ZA | train,test |
232
+ | venda and xitsonga | 188 | 2 | 2603.75 | ZA | train,test |
233
+ | sotho | 182 | 4 | 2082.21 | ZA | dev,test,train |
234
+ | akan | 157 | 6 | 1392.47 | GH | test,train |
235
+ | nupe | 156 | 9 | 1608.24 | NG | dev,train,test |
236
+ | anaang | 153 | 8 | 1532.56 | NG | test,dev |
237
+ | english | 151 | 11 | 2445.98 | NG | dev,test |
238
+ | afemai | 142 | 2 | 1877.04 | NG | train,test |
239
+ | shona | 138 | 8 | 1419.98 | ZA,ZW | test,train,dev |
240
+ | eggon | 137 | 5 | 1833.77 | NG | test |
241
+ | luganda and kiswahili | 134 | 1 | 1356.93 | UG | train |
242
+ | ukwuani | 133 | 7 | 1269.02 | NG | test |
243
+ | sesotho | 132 | 10 | 1397.16 | ZA | train,dev,test |
244
+ | benin | 124 | 4 | 1457.48 | NG | train,test |
245
+ | kagoma | 123 | 1 | 1781.04 | NG | train |
246
+ | nasarawa eggon | 120 | 1 | 1039.99 | NG | train |
247
+ | tiv | 120 | 14 | 1084.52 | NG | train,test,dev |
248
+ | south african english | 119 | 2 | 1643.82 | ZA | train,test |
249
+ | borana | 112 | 1 | 1090.71 | KE | train |
250
+ | swahili ,luganda ,arabic | 109 | 1 | 929.46 | UG | train |
251
+ | ogoni | 109 | 4 | 1629.7 | NG | train,test |
252
+ | mada | 109 | 2 | 1786.26 | NG | test |
253
+ | bette | 106 | 4 | 930.16 | NG | train,test |
254
+ | berom | 105 | 4 | 1272.99 | NG | dev,test |
255
+ | bini | 104 | 4 | 1499.75 | NG | test |
256
+ | ngas | 102 | 3 | 1234.16 | NG | train,test |
257
+ | etsako | 101 | 4 | 1074.53 | NG | train,test |
258
+ | okrika | 100 | 3 | 1887.47 | NG | train,test |
259
+ | venda | 99 | 2 | 938.14 | ZA | train,test |
260
+ | siswati | 96 | 5 | 1367.45 | ZA | dev,train,test |
261
+ | damara | 92 | 1 | 674.43 | NG | train |
262
+ | yoruba, hausa | 89 | 5 | 928.98 | NG | test |
263
+ | southern sotho | 89 | 1 | 889.73 | ZA | train |
264
+ | kanuri | 86 | 7 | 1936.78 | NG | test,dev |
265
+ | itsekiri | 82 | 3 | 778.47 | NG | test,dev |
266
+ | ekpeye | 80 | 2 | 922.88 | NG | test |
267
+ | mwaghavul | 78 | 2 | 738.02 | NG | test |
268
+ | bajju | 72 | 2 | 758.16 | NG | test |
269
+ | luo, swahili | 71 | 1 | 616.57 | KE | train |
270
+ | dholuo | 70 | 1 | 669.07 | KE | train |
271
+ | ekene | 68 | 1 | 839.31 | NG | test |
272
+ | jaba | 65 | 2 | 540.66 | NG | test |
273
+ | ika | 65 | 4 | 576.56 | NG | test,dev |
274
+ | angas | 65 | 1 | 589.99 | NG | test |
275
+ | ateso | 63 | 1 | 624.28 | UG | train |
276
+ | brass | 62 | 2 | 900.04 | NG | test |
277
+ | ikulu | 61 | 1 | 313.2 | NG | test |
278
+ | eleme | 60 | 2 | 1207.92 | NG | test |
279
+ | chichewa | 60 | 1 | 554.61 | MW | train |
280
+ | oklo | 58 | 1 | 871.37 | NG | test |
281
+ | meru | 58 | 2 | 865.07 | KE | train,test |
282
+ | agatu | 55 | 1 | 369.11 | NG | test |
283
+ | okirika | 54 | 1 | 792.65 | NG | test |
284
+ | igarra | 54 | 1 | 562.12 | NG | test |
285
+ | ijaw(nembe) | 54 | 2 | 537.56 | NG | test |
286
+ | khana | 51 | 2 | 497.42 | NG | test |
287
+ | ogbia | 51 | 4 | 461.15 | NG | test,dev |
288
+ | gbagyi | 51 | 4 | 693.43 | NG | test |
289
+ | portuguese | 50 | 1 | 525.02 | ZA | train |
290
+ | delta | 49 | 2 | 425.76 | NG | test |
291
+ | bassa | 49 | 1 | 646.13 | NG | test |
292
+ | etche | 49 | 1 | 637.48 | NG | test |
293
+ | kubi | 46 | 1 | 495.21 | NG | test |
294
+ | jukun | 44 | 2 | 362.12 | NG | test |
295
+ | igbo and yoruba | 43 | 2 | 466.98 | NG | test |
296
+ | urobo | 43 | 3 | 573.14 | NG | test |
297
+ | kalabari | 42 | 5 | 305.49 | NG | test |
298
+ | ibani | 42 | 1 | 322.34 | NG | test |
299
+ | obolo | 37 | 1 | 204.79 | NG | test |
300
+ | idah | 34 | 1 | 533.5 | NG | test |
301
+ | bassa-nge/nupe | 31 | 3 | 267.42 | NG | test,dev |
302
+ | yala mbembe | 29 | 1 | 237.27 | NG | test |
303
+ | eket | 28 | 1 | 238.85 | NG | test |
304
+ | afo | 26 | 1 | 171.15 | NG | test |
305
+ | ebiobo | 25 | 1 | 226.27 | NG | test |
306
+ | nyandang | 25 | 1 | 230.41 | NG | test |
307
+ | ishan | 23 | 1 | 194.12 | NG | test |
308
+ | bagi | 20 | 1 | 284.54 | NG | test |
309
+ | estako | 20 | 1 | 480.78 | NG | test |
310
+ | gerawa | 13 | 1 | 342.15 | NG | test |
311
+
312
+ #### Initial Data Collection and Normalization
313
+
314
+ [Needs More Information]
315
+
316
+ #### Who are the source language producers?
317
+
318
+ [Needs More Information]
319
+
320
+ ### Annotations
321
+
322
+ #### Annotation process
323
+
324
+ [Needs More Information]
325
+
326
+ #### Who are the annotators?
327
+
328
+ [Needs More Information]
329
+
330
+ ### Personal and Sensitive Information
331
+
332
+ The dataset consists of people who have donated their voice online. You agree to not attempt to determine the identity of speakers in this dataset.
333
+
334
+ ## Considerations for Using the Data
335
+
336
+ ### Social Impact of Dataset
337
+
338
+ [More Information Needed]
339
+
340
+ ### Discussion of Biases
341
+
342
+ [More Information Needed]
343
+
344
+ ### Other Known Limitations
345
+
346
+ Dataset provided for research purposes only. Please check dataset license for additional information.
347
+
348
+ ## Additional Information
349
+
350
+ ### Dataset Curators
351
+
352
+ The dataset was initially prepared by Intron and refined for public release by CLAIR Lab.
353
+
354
+ ### Licensing Information
355
+
356
+ Public Domain, Creative Commons Attribution NonCommercial ShareAlike v4.0 ([CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode))
357
+
358
+ ### Citation Information
359
+
360
+ [More Information Needed]
361
+
362
+ ### Contributions
363
+
364
+ Thanks to [@tobiolatunji](https://github.com/tobiolatunji) for adding this dataset.
afrispeech-200.py ADDED
@@ -0,0 +1,160 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2023 The HuggingFace Datasets Authors and the current dataset script contributor.
3
+ #
4
+ # Licensed under the Apache License, Version 2.0 (the "License");
5
+ # you may not use this file except in compliance with the License.
6
+ # You may obtain a copy of the License at
7
+ #
8
+ # http://www.apache.org/licenses/LICENSE-2.0
9
+ #
10
+ # Unless required by applicable law or agreed to in writing, software
11
+ # distributed under the License is distributed on an "AS IS" BASIS,
12
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13
+ # See the License for the specific language governing permissions and
14
+ # limitations under the License.
15
+ """ AfriSpeech-200 Dataset"""
16
+
17
+
18
+ import csv
19
+ import os
20
+
21
+ import datasets
22
+ from datasets.utils.py_utils import size_str
23
+ from tqdm import tqdm
24
+
25
+
26
+ _CITATION = """ TBD """
27
+
28
+ _DESCRIPTION = """\
29
+ AFRISPEECH-200 is a 200hr Pan-African speech corpus for clinical and general domain English accented ASR;
30
+ a dataset with 120 African accents from 13 countries and 2,463 unique African speakers.
31
+ Our goal is to raise awareness for and advance Pan-African English ASR research,
32
+ especially for the clinical domain.
33
+ """
34
+
35
+ _HOMEPAGE = "https://github.com/intron-innovation/AfriSpeech-Dataset-Paper"
36
+
37
+ _LICENSE = "http://creativecommons.org/licenses/by-nc-sa/4.0/"
38
+
39
+
40
+ # TODO: change "streaming" to "main" after merge!
41
+ _BASE_URL = "https://huggingface.co/datasets/intron/afrispeech-200/main/"
42
+
43
+ _AUDIO_URL = _BASE_URL + "audio/{split}/{split}_{shard_idx}.tar.gz"
44
+
45
+ _TRANSCRIPT_URL = _BASE_URL + "transcripts/{split}.csv"
46
+
47
+ _SHARDS = {
48
+ 'train': 35,
49
+ 'dev': 2,
50
+ 'test': 4
51
+ }
52
+
53
+
54
+ class AfriSpeech(datasets.GeneratorBasedBuilder):
55
+ DEFAULT_WRITER_BATCH_SIZE = 1000
56
+ VERSION = datasets.Version("1.1.0")
57
+ DEFAULT_CONFIG_NAME = "all"
58
+
59
+ def _info(self):
60
+ description = _DESCRIPTION
61
+ features = datasets.Features(
62
+ {
63
+ "speaker_id": datasets.Value("string"),
64
+ "path": datasets.Value("string"),
65
+ "audio": datasets.features.Audio(sampling_rate=44_100),
66
+ "transcript": datasets.Value("string"),
67
+ "age_group": datasets.Value("string"),
68
+ "gender": datasets.Value("string"),
69
+ "accent": datasets.Value("string"),
70
+ "domain": datasets.Value("string"),
71
+ "country": datasets.Value("string"),
72
+ "duration": datasets.Value("float"),
73
+ }
74
+ )
75
+
76
+ return datasets.DatasetInfo(
77
+ description=description,
78
+ features=features,
79
+ supervised_keys=None,
80
+ homepage=_HOMEPAGE,
81
+ license=_LICENSE,
82
+ citation=_CITATION,
83
+ version=self.VERSION,
84
+ )
85
+
86
+ def _split_generators(self, dl_manager):
87
+ """Returns SplitGenerators."""
88
+ # If several configurations are possible (listed in BUILDER_CONFIGS),
89
+ # the configuration selected by the user is in self.config.name
90
+
91
+ # dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLs
92
+ # It can accept any type or nested list/dict and will give back the same structure
93
+ # with the url replaced with path to local files.
94
+ # By default the archives will be extracted and a path to a cached folder
95
+ # where they are extracted is returned instead of the archive
96
+
97
+ n_shards = _SHARDS
98
+
99
+ audio_urls = {}
100
+ splits = ("train", "dev") # , "test"
101
+ for split in splits:
102
+ audio_urls[split] = [
103
+ _AUDIO_URL.format(split=split, shard_idx=i) for i in range(n_shards[split])
104
+ ]
105
+ archive_paths = dl_manager.download(audio_urls)
106
+ local_extracted_archive_paths = dl_manager.extract(archive_paths) if not dl_manager.is_streaming else {}
107
+
108
+ meta_urls = {split: _TRANSCRIPT_URL.format(split=split) for split in splits}
109
+ meta_paths = dl_manager.download_and_extract(meta_urls)
110
+
111
+ split_generators = []
112
+ split_names = {
113
+ "train": datasets.Split.TRAIN,
114
+ "dev": datasets.Split.VALIDATION,
115
+ # "test": datasets.Split.TEST,
116
+ }
117
+ for split in splits:
118
+ split_generators.append(
119
+ datasets.SplitGenerator(
120
+ name=split_names.get(split, split),
121
+ gen_kwargs={
122
+ "local_extracted_archive_paths": local_extracted_archive_paths.get(split),
123
+ "archives": [dl_manager.iter_archive(path) for path in archive_paths.get(split)],
124
+ "meta_path": meta_paths[split],
125
+ },
126
+ ),
127
+ )
128
+
129
+ return split_generators
130
+
131
+ def _generate_examples(self, local_extracted_archive_paths, archives, meta_path):
132
+ """Yields examples as (key, example) tuples."""
133
+ # This method handles input defined in _split_generators to yield (key, example) tuples
134
+ # from the dataset.
135
+ # The `key` is here for legacy reason (tfds) and is not important in itself.
136
+
137
+ data_fields = list(self._info().features.keys())
138
+ metadata = {}
139
+ with open(meta_path, "r", encoding="utf-8") as f:
140
+ reader = csv.DictReader(f)
141
+ for row in tqdm(reader, desc="Reading metadata..."):
142
+ row["speaker_id"] = row["user_ids"]
143
+ # if data is incomplete, fill with empty values
144
+ for field in data_fields:
145
+ if field not in row:
146
+ row[field] = ""
147
+ metadata[row["audio_paths"]] = row
148
+
149
+ for i, audio_archive in enumerate(archives):
150
+ for filename, file in audio_archive:
151
+ _, filename = os.path.split(filename)
152
+ if filename in metadata:
153
+ result = dict(metadata[filename])
154
+ # set the audio feature and the path to the extracted file
155
+ path = os.path.join(local_extracted_archive_paths[i], filename) if local_extracted_archive_paths else filename
156
+ result["audio"] = {"path": path, "bytes": file.read()}
157
+ # set path to None if the audio file doesn't exist locally (i.e. in streaming mode)
158
+ result["path"] = path if local_extracted_archive_paths else filename
159
+
160
+ yield path, result
audio/dev/dev_0.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad5799fabc150c6489af97c80de393ad0d3f841dd64eb057d92fb0acf4972f78
3
+ size 1444447165
audio/dev/dev_1.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2dc5abbffcc23b45113167b208bd1a19c56fcc59fc298e5e220e47b236810d82
3
+ size 1251443845
audio/test/test_0.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c273387f4f047cde01c705d71fd149c92dde793219917262877af0228d3c1600
3
+ size 1343919705
audio/test/test_1.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a1e20d419f4a6724945a4976d479af59e475da81626d148cbc4509b76c28ca5f
3
+ size 1384883073
audio/test/test_2.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:73e1b861ff25427e6909820306d6db6233a0671be5a636c643041274d8e67f20
3
+ size 1654173614
audio/test/test_3.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a8e0d2fd5fc86d7e611c882b016188c22c8a2a487043989c53e0c8bb737d43e
3
+ size 1157250172
audio/train/train_0.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b847e262a4c180af2e74d5bfcfe14953caae80b489604fec44c39abe296daa9
3
+ size 1329256779
audio/train/train_1.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c4dac1ba79c8a20d57be8a784860467b5abd3a373f5ee637386cfd2197a26b5a
3
+ size 1333937266
audio/train/train_10.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ce10933bf2a0b9882914191d3a57a190c5c3b07649057a46b0baacd4686a7ea
3
+ size 1540014632
audio/train/train_11.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e93a2a6dccc9eed09b2408947eeee3a83812c294d531e63009662780806b77b0
3
+ size 1556482774
audio/train/train_12.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d547c5274e177d90d0e540f15029e392b55040edca10cf5449eeb975bbba2b3
3
+ size 1589868802
audio/train/train_13.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:071ec4b4a2af1c0e02f73b518fddd0b2bec826c72f463ac2f79c8fccad928796
3
+ size 1583525018
audio/train/train_14.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e77986f86262155e0c0da9dbb5d195c9a99da489f44995656dcd61288975dcd8
3
+ size 1586219707
audio/train/train_15.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a492feb6870b2560e06a87ffcf40ddc78fd94b76450deec98521e1a6ea745201
3
+ size 1601044578
audio/train/train_16.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb0b78cb6186a0ac0807a13f66549a68dba08e6d11364e895136033fb460a52d
3
+ size 1585414449
audio/train/train_17.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ce04280b39925ff1cb4f75206f2fa3761209a3acaa5542bd1fe2117c95ea664
3
+ size 1564635736
audio/train/train_18.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d16c00885892f70c01f6cc5c5b1ec62cbb60f88cb250d6214cca113925e3e60
3
+ size 1570527240
audio/train/train_19.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9010ddb110a222cfc368fd2f4ca644521981aabb888044bf898ac0f036db3468
3
+ size 1544645646
audio/train/train_2.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b8205f254b3148337c8e09c58ea1ec04f95d097206346643aea53b83dc2c759
3
+ size 1309380968
audio/train/train_20.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18166c5fd676bdcaf386828f9132d0ac96794e53026d0885ba81e9e2881ff863
3
+ size 1597562952
audio/train/train_21.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f34c2db602294e3017fc405b61767e3eb3f5f819fa3764d17d2f1d682e79353
3
+ size 1592584723
audio/train/train_22.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79d3f7490a7fb71597ea6ebf3bb4ccd9975888ac78a1e091f2f6bf6f4a080e53
3
+ size 1548080538
audio/train/train_23.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:951d05c3232fdb70cac4bfef8e7c7bd72fe690fdc943306ffb9d68f657d1a02f
3
+ size 1564081923
audio/train/train_24.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:20850503d98a1ce776b0b2fddbca998c982881a245f7582fb694bf70c9258b4a
3
+ size 1832469461
audio/train/train_25.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5675992efb0e667063bbd35b4b17f68edf8c8d8ad837033b222eccfbe156f19
3
+ size 1763814490
audio/train/train_26.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f741e7e2e92b8c7b024dc7db8f7b17b17489ce86e1ea008aadcc407edd7616aa
3
+ size 1807204833
audio/train/train_27.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f0230deb1f7a6b2a330d072f53f75eaea77948edd10811788aadb05c9b8f093d
3
+ size 1859470321
audio/train/train_28.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6bfb51c0a5093c3d727361e20395bf13a5049a76012ff59907a97974181c4c85
3
+ size 1795748605
audio/train/train_29.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e705fc4d025cb1d0953f9ae4fbb093bd90ec4c71b2e5a6c50cf9466a35212120
3
+ size 1514953567
audio/train/train_3.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b26cea54e5397ef63f3f8553613845e71ff5703d1ed843168de6cdf5a221c76
3
+ size 1417376892
audio/train/train_30.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:452231f2295f260888b0a3b8f0a02ac00fc57983e099f9727a5a21fe8afdaa98
3
+ size 1560222270
audio/train/train_31.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:900ce87d3c826c907984110735a0d697d58d8c67d64de5d11d5507007307a7ac
3
+ size 1595039903
audio/train/train_32.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e941d4a6202f46b349b5c7574dbb8574731f98cc42afc5dfefa573b7378cdef6
3
+ size 1714877917
audio/train/train_33.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:963ae09353086b2d2b19706977a2de0f5fcd59a3c237c88928a9effc1c878834
3
+ size 1569196965
audio/train/train_34.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:55df0fb43b055873f8f94d56f827d5b3036b50cdea14d8817a09554f422260c2
3
+ size 482454366
audio/train/train_4.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e89d6ae9f39efceeb622020d3ec0922181d3a9e3331749e66cf4d4ec7ee8aa0
3
+ size 1497037115
audio/train/train_5.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4411f65c0a7f505c5f3a7b062654a66698c5dc8746ed1b6de6dadcc7365b6aa6
3
+ size 1444022238
audio/train/train_6.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce739d36bef48597b5ba6b5c6c95717fa1038ec80cbf9174bbf93bb94972157b
3
+ size 1464217532
audio/train/train_7.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c92c48d96830a12c89b96bd1a56bfcf111ec08e0eeb654814f48ae1c5e0a0b7
3
+ size 1595095584
audio/train/train_8.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f407ce9432bc68854f890ccc368bc1a4561cf726cce18f8202e6f95340415aac
3
+ size 1456457665
audio/train/train_9.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c7e1245d80dbf0797c3fbc8f6be5f48e0a609ec397266fdf5615070560058bd
3
+ size 1463734308
transcripts/dev.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bfc0cb4edfd2809bdc0959a55c12f49bfa72317e1d1e290cc6197cbcd011336e
3
+ size 1079417
transcripts/test.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa7699d9428e97d904d9de0502f0abe265d54daaf09df7f97657558ee83f834f
3
+ size 2111887