Update README.md
Browse files
README.md
CHANGED
@@ -12,6 +12,7 @@ language:
|
|
12 |
- es
|
13 |
- pt
|
14 |
- pl
|
|
|
15 |
license:
|
16 |
- cc-by-4.0
|
17 |
multilinguality:
|
@@ -22,6 +23,8 @@ source_datasets:
|
|
22 |
- original
|
23 |
task_categories:
|
24 |
- automatic-speech-recognition
|
|
|
|
|
25 |
paperswithcode_id: multilingual-librispeech
|
26 |
pretty_name: MultiLingual LibriSpeech
|
27 |
dataset_info:
|
@@ -49,19 +52,19 @@ dataset_info:
|
|
49 |
dtype: string
|
50 |
splits:
|
51 |
- name: dev
|
52 |
-
num_bytes: 199959986
|
53 |
num_examples: 3095
|
54 |
- name: test
|
55 |
-
num_bytes: 199298575
|
56 |
num_examples: 3075
|
57 |
- name: train
|
58 |
-
num_bytes: 23931679031
|
59 |
num_examples: 374287
|
60 |
- name: 9_hours
|
61 |
num_bytes: 139884664.668
|
62 |
num_examples: 2153
|
63 |
- name: 1_hours
|
64 |
-
num_bytes: 15462181
|
65 |
num_examples: 234
|
66 |
download_size: 24376256629
|
67 |
dataset_size: 24486284437.668
|
@@ -101,7 +104,7 @@ dataset_info:
|
|
101 |
num_bytes: 142796680.609
|
102 |
num_examples: 2167
|
103 |
- name: 1_hours
|
104 |
-
num_bytes: 15675831
|
105 |
num_examples: 241
|
106 |
download_size: 17381581776
|
107 |
dataset_size: 17459684482.927002
|
@@ -135,13 +138,13 @@ dataset_info:
|
|
135 |
num_bytes: 225756069.096
|
136 |
num_examples: 3394
|
137 |
- name: train
|
138 |
-
num_bytes: 31050881388
|
139 |
num_examples: 469942
|
140 |
- name: 9_hours
|
141 |
num_bytes: 142777983.118
|
142 |
num_examples: 2194
|
143 |
- name: 1_hours
|
144 |
-
num_bytes: 15714704
|
145 |
num_examples: 241
|
146 |
download_size: 31526161821
|
147 |
dataset_size: 31659423725.516
|
@@ -175,13 +178,13 @@ dataset_info:
|
|
175 |
num_bytes: 83216752.046
|
176 |
num_examples: 1262
|
177 |
- name: train
|
178 |
-
num_bytes: 3896742625
|
179 |
num_examples: 59623
|
180 |
- name: 9_hours
|
181 |
num_bytes: 141671904.428
|
182 |
num_examples: 2173
|
183 |
- name: 1_hours
|
184 |
-
num_bytes: 15560398
|
185 |
num_examples: 240
|
186 |
download_size: 4200633596
|
187 |
dataset_size: 4218799275.522
|
@@ -209,22 +212,22 @@ dataset_info:
|
|
209 |
dtype: string
|
210 |
splits:
|
211 |
- name: dev
|
212 |
-
num_bytes: 32746725
|
213 |
num_examples: 512
|
214 |
- name: test
|
215 |
-
num_bytes: 33735044
|
216 |
num_examples: 520
|
217 |
- name: train
|
218 |
-
num_bytes: 1638889846
|
219 |
num_examples: 25043
|
220 |
- name: 9_hours
|
221 |
-
num_bytes: 142005461
|
222 |
num_examples: 2173
|
223 |
- name: 1_hours
|
224 |
-
num_bytes: 15681216
|
225 |
num_examples: 238
|
226 |
download_size: 1855342312
|
227 |
-
dataset_size: 1863058292
|
228 |
- config_name: portuguese
|
229 |
features:
|
230 |
- name: audio
|
@@ -249,10 +252,10 @@ dataset_info:
|
|
249 |
dtype: string
|
250 |
splits:
|
251 |
- name: dev
|
252 |
-
num_bytes: 57533473
|
253 |
num_examples: 826
|
254 |
- name: test
|
255 |
-
num_bytes: 59141979
|
256 |
num_examples: 871
|
257 |
- name: train
|
258 |
num_bytes: 2518553713.946
|
@@ -261,7 +264,7 @@ dataset_info:
|
|
261 |
num_bytes: 141641902.42
|
262 |
num_examples: 2116
|
263 |
- name: 1_hours
|
264 |
-
num_bytes: 15697139
|
265 |
num_examples: 236
|
266 |
download_size: 2780836500
|
267 |
dataset_size: 2792568207.366
|
@@ -295,13 +298,13 @@ dataset_info:
|
|
295 |
num_bytes: 158526899.32
|
296 |
num_examples: 2385
|
297 |
- name: train
|
298 |
-
num_bytes: 14562584188
|
299 |
num_examples: 220701
|
300 |
- name: 9_hours
|
301 |
num_bytes: 142473624.48
|
302 |
num_examples: 2110
|
303 |
- name: 1_hours
|
304 |
-
num_bytes: 15702048
|
305 |
num_examples: 233
|
306 |
download_size: 14971394533
|
307 |
dataset_size: 15037091662.944
|
@@ -432,11 +435,12 @@ This is a streamable version of the Multilingual LibriSpeech (MLS) dataset.
|
|
432 |
The data archives were restructured from the original ones from [OpenSLR](http://www.openslr.org/94) to make it easier to stream.
|
433 |
|
434 |
MLS dataset is a large multilingual corpus suitable for speech research. The dataset is derived from read audiobooks from LibriVox and consists of
|
435 |
-
8 languages - English, German, Dutch, Spanish, French, Italian, Portuguese, Polish.
|
436 |
|
437 |
### Supported Tasks and Leaderboards
|
438 |
|
439 |
- `automatic-speech-recognition`, `speaker-identification`: The dataset can be used to train a model for Automatic Speech Recognition (ASR). The model is presented with an audio file and asked to transcribe the audio file to written text. The most common evaluation metric is the word error rate (WER). The task has an active leaderboard which can be found at https://paperswithcode.com/dataset/multilingual-librispeech and ranks models based on their WER.
|
|
|
440 |
|
441 |
### Languages
|
442 |
|
@@ -449,16 +453,13 @@ The `datasets` library allows you to load and pre-process your dataset in pure P
|
|
449 |
For example, to download the German config, simply specify the corresponding language config name (i.e., "german" for German):
|
450 |
```python
|
451 |
from datasets import load_dataset
|
452 |
-
|
453 |
mls = load_dataset("facebook/multilingual_librispeech", "german", split="train")
|
454 |
```
|
455 |
|
456 |
Using the datasets library, you can also stream the dataset on-the-fly by adding a `streaming=True` argument to the `load_dataset` function call. Loading a dataset in streaming mode loads individual samples of the dataset at a time, rather than downloading the entire dataset to disk.
|
457 |
```python
|
458 |
from datasets import load_dataset
|
459 |
-
|
460 |
mls = load_dataset("facebook/multilingual_librispeech", "german", split="train", streaming=True)
|
461 |
-
|
462 |
print(next(iter(mls)))
|
463 |
```
|
464 |
|
@@ -469,7 +470,6 @@ Local:
|
|
469 |
```python
|
470 |
from datasets import load_dataset
|
471 |
from torch.utils.data.sampler import BatchSampler, RandomSampler
|
472 |
-
|
473 |
mls = load_dataset("facebook/multilingual_librispeech", "german", split="train")
|
474 |
batch_sampler = BatchSampler(RandomSampler(mls), batch_size=32, drop_last=False)
|
475 |
dataloader = DataLoader(mls, batch_sampler=batch_sampler)
|
@@ -480,7 +480,6 @@ Streaming:
|
|
480 |
```python
|
481 |
from datasets import load_dataset
|
482 |
from torch.utils.data import DataLoader
|
483 |
-
|
484 |
mls = load_dataset("facebook/multilingual_librispeech", "german", split="train", streaming=True)
|
485 |
dataloader = DataLoader(mls, batch_size=32)
|
486 |
```
|
@@ -521,12 +520,11 @@ A typical data point comprises the path to the audio file, usually called `file`
|
|
521 |
- id: unique id of the data sample.
|
522 |
|
523 |
- speaker_id: unique id of the speaker. The same speaker id can be found for multiple data samples.
|
524 |
-
|
525 |
- chapter_id: id of the audiobook chapter which includes the transcription.
|
526 |
|
527 |
### Data Splits
|
528 |
|
529 |
-
|
|
530 |
| ----- | ------ | ----- | ---- | ---- | ---- |
|
531 |
| german | 469942 | 2194 | 241 | 3469 | 3394 |
|
532 |
| dutch | 374287 | 2153 | 234 | 3095 | 3075 |
|
@@ -536,8 +534,6 @@ A typical data point comprises the path to the audio file, usually called `file`
|
|
536 |
| portuguese | 37533 | 2116 | 236 | 826 | 871 |
|
537 |
| polish | 25043 | 2173 | 238 | 512 | 520 |
|
538 |
|
539 |
-
|
540 |
-
|
541 |
## Dataset Creation
|
542 |
|
543 |
### Curation Rationale
|
@@ -604,7 +600,47 @@ Public Domain, Creative Commons Attribution 4.0 International Public License ([C
|
|
604 |
}
|
605 |
```
|
606 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
607 |
### Contributions
|
608 |
|
609 |
-
Thanks to [@patrickvonplaten](https://github.com/patrickvonplaten)
|
610 |
-
and [@polinaeterna](https://github.com/polinaeterna) for adding this dataset.
|
|
|
12 |
- es
|
13 |
- pt
|
14 |
- pl
|
15 |
+
- en
|
16 |
license:
|
17 |
- cc-by-4.0
|
18 |
multilinguality:
|
|
|
23 |
- original
|
24 |
task_categories:
|
25 |
- automatic-speech-recognition
|
26 |
+
- text-to-speech
|
27 |
+
- text-to-audio
|
28 |
paperswithcode_id: multilingual-librispeech
|
29 |
pretty_name: MultiLingual LibriSpeech
|
30 |
dataset_info:
|
|
|
52 |
dtype: string
|
53 |
splits:
|
54 |
- name: dev
|
55 |
+
num_bytes: 199959986
|
56 |
num_examples: 3095
|
57 |
- name: test
|
58 |
+
num_bytes: 199298575
|
59 |
num_examples: 3075
|
60 |
- name: train
|
61 |
+
num_bytes: 23931679031
|
62 |
num_examples: 374287
|
63 |
- name: 9_hours
|
64 |
num_bytes: 139884664.668
|
65 |
num_examples: 2153
|
66 |
- name: 1_hours
|
67 |
+
num_bytes: 15462181
|
68 |
num_examples: 234
|
69 |
download_size: 24376256629
|
70 |
dataset_size: 24486284437.668
|
|
|
104 |
num_bytes: 142796680.609
|
105 |
num_examples: 2167
|
106 |
- name: 1_hours
|
107 |
+
num_bytes: 15675831
|
108 |
num_examples: 241
|
109 |
download_size: 17381581776
|
110 |
dataset_size: 17459684482.927002
|
|
|
138 |
num_bytes: 225756069.096
|
139 |
num_examples: 3394
|
140 |
- name: train
|
141 |
+
num_bytes: 31050881388
|
142 |
num_examples: 469942
|
143 |
- name: 9_hours
|
144 |
num_bytes: 142777983.118
|
145 |
num_examples: 2194
|
146 |
- name: 1_hours
|
147 |
+
num_bytes: 15714704
|
148 |
num_examples: 241
|
149 |
download_size: 31526161821
|
150 |
dataset_size: 31659423725.516
|
|
|
178 |
num_bytes: 83216752.046
|
179 |
num_examples: 1262
|
180 |
- name: train
|
181 |
+
num_bytes: 3896742625
|
182 |
num_examples: 59623
|
183 |
- name: 9_hours
|
184 |
num_bytes: 141671904.428
|
185 |
num_examples: 2173
|
186 |
- name: 1_hours
|
187 |
+
num_bytes: 15560398
|
188 |
num_examples: 240
|
189 |
download_size: 4200633596
|
190 |
dataset_size: 4218799275.522
|
|
|
212 |
dtype: string
|
213 |
splits:
|
214 |
- name: dev
|
215 |
+
num_bytes: 32746725
|
216 |
num_examples: 512
|
217 |
- name: test
|
218 |
+
num_bytes: 33735044
|
219 |
num_examples: 520
|
220 |
- name: train
|
221 |
+
num_bytes: 1638889846
|
222 |
num_examples: 25043
|
223 |
- name: 9_hours
|
224 |
+
num_bytes: 142005461
|
225 |
num_examples: 2173
|
226 |
- name: 1_hours
|
227 |
+
num_bytes: 15681216
|
228 |
num_examples: 238
|
229 |
download_size: 1855342312
|
230 |
+
dataset_size: 1863058292
|
231 |
- config_name: portuguese
|
232 |
features:
|
233 |
- name: audio
|
|
|
252 |
dtype: string
|
253 |
splits:
|
254 |
- name: dev
|
255 |
+
num_bytes: 57533473
|
256 |
num_examples: 826
|
257 |
- name: test
|
258 |
+
num_bytes: 59141979
|
259 |
num_examples: 871
|
260 |
- name: train
|
261 |
num_bytes: 2518553713.946
|
|
|
264 |
num_bytes: 141641902.42
|
265 |
num_examples: 2116
|
266 |
- name: 1_hours
|
267 |
+
num_bytes: 15697139
|
268 |
num_examples: 236
|
269 |
download_size: 2780836500
|
270 |
dataset_size: 2792568207.366
|
|
|
298 |
num_bytes: 158526899.32
|
299 |
num_examples: 2385
|
300 |
- name: train
|
301 |
+
num_bytes: 14562584188
|
302 |
num_examples: 220701
|
303 |
- name: 9_hours
|
304 |
num_bytes: 142473624.48
|
305 |
num_examples: 2110
|
306 |
- name: 1_hours
|
307 |
+
num_bytes: 15702048
|
308 |
num_examples: 233
|
309 |
download_size: 14971394533
|
310 |
dataset_size: 15037091662.944
|
|
|
435 |
The data archives were restructured from the original ones from [OpenSLR](http://www.openslr.org/94) to make it easier to stream.
|
436 |
|
437 |
MLS dataset is a large multilingual corpus suitable for speech research. The dataset is derived from read audiobooks from LibriVox and consists of
|
438 |
+
8 languages - English, German, Dutch, Spanish, French, Italian, Portuguese, Polish. It includes about 44.5K hours of English and a total of about 6K hours for other languages.
|
439 |
|
440 |
### Supported Tasks and Leaderboards
|
441 |
|
442 |
- `automatic-speech-recognition`, `speaker-identification`: The dataset can be used to train a model for Automatic Speech Recognition (ASR). The model is presented with an audio file and asked to transcribe the audio file to written text. The most common evaluation metric is the word error rate (WER). The task has an active leaderboard which can be found at https://paperswithcode.com/dataset/multilingual-librispeech and ranks models based on their WER.
|
443 |
+
- `text-to-speech`, `text-to-audio`: The dataset can also be used to train a model for Text-To-Speech (TTS).
|
444 |
|
445 |
### Languages
|
446 |
|
|
|
453 |
For example, to download the German config, simply specify the corresponding language config name (i.e., "german" for German):
|
454 |
```python
|
455 |
from datasets import load_dataset
|
|
|
456 |
mls = load_dataset("facebook/multilingual_librispeech", "german", split="train")
|
457 |
```
|
458 |
|
459 |
Using the datasets library, you can also stream the dataset on-the-fly by adding a `streaming=True` argument to the `load_dataset` function call. Loading a dataset in streaming mode loads individual samples of the dataset at a time, rather than downloading the entire dataset to disk.
|
460 |
```python
|
461 |
from datasets import load_dataset
|
|
|
462 |
mls = load_dataset("facebook/multilingual_librispeech", "german", split="train", streaming=True)
|
|
|
463 |
print(next(iter(mls)))
|
464 |
```
|
465 |
|
|
|
470 |
```python
|
471 |
from datasets import load_dataset
|
472 |
from torch.utils.data.sampler import BatchSampler, RandomSampler
|
|
|
473 |
mls = load_dataset("facebook/multilingual_librispeech", "german", split="train")
|
474 |
batch_sampler = BatchSampler(RandomSampler(mls), batch_size=32, drop_last=False)
|
475 |
dataloader = DataLoader(mls, batch_sampler=batch_sampler)
|
|
|
480 |
```python
|
481 |
from datasets import load_dataset
|
482 |
from torch.utils.data import DataLoader
|
|
|
483 |
mls = load_dataset("facebook/multilingual_librispeech", "german", split="train", streaming=True)
|
484 |
dataloader = DataLoader(mls, batch_size=32)
|
485 |
```
|
|
|
520 |
- id: unique id of the data sample.
|
521 |
|
522 |
- speaker_id: unique id of the speaker. The same speaker id can be found for multiple data samples.
|
|
|
523 |
- chapter_id: id of the audiobook chapter which includes the transcription.
|
524 |
|
525 |
### Data Splits
|
526 |
|
527 |
+
| Number of samples | Train | Train.9h | Train.1h | Dev | Test |
|
528 |
| ----- | ------ | ----- | ---- | ---- | ---- |
|
529 |
| german | 469942 | 2194 | 241 | 3469 | 3394 |
|
530 |
| dutch | 374287 | 2153 | 234 | 3095 | 3075 |
|
|
|
534 |
| portuguese | 37533 | 2116 | 236 | 826 | 871 |
|
535 |
| polish | 25043 | 2173 | 238 | 512 | 520 |
|
536 |
|
|
|
|
|
537 |
## Dataset Creation
|
538 |
|
539 |
### Curation Rationale
|
|
|
600 |
}
|
601 |
```
|
602 |
|
603 |
+
|
604 |
+
### Data Statistics
|
605 |
+
|
606 |
+
| Duration (h) | Train | Dev | Test |
|
607 |
+
|--------------|-----------|-------|-------|
|
608 |
+
| English | 44,659.74 | 15.75 | 15.55 |
|
609 |
+
| German | 1,966.51 | 14.28 | 14.29 |
|
610 |
+
| Dutch | 1,554.24 | 12.76 | 12.76 |
|
611 |
+
| French | 1,076.58 | 10.07 | 10.07 |
|
612 |
+
| Spanish | 917.68 | 9.99 | 10 |
|
613 |
+
| Italian | 247.38 | 5.18 | 5.27 |
|
614 |
+
| Portuguese | 160.96 | 3.64 | 3.74 |
|
615 |
+
| Polish | 103.65 | 2.08 | 2.14 |
|
616 |
+
|
617 |
+
| # Speakers | Train | | Dev | | Test | |
|
618 |
+
|------------|-------|------|-----|----|------|----|
|
619 |
+
| Gender | M | F | M | F | M | F |
|
620 |
+
| English | 2742 | 2748 | 21 | 21 | 21 | 21 |
|
621 |
+
| German | 81 | 95 | 15 | 15 | 15 | 15 |
|
622 |
+
| Dutch | 9 | 31 | 3 | 3 | 3 | 3 |
|
623 |
+
| French | 62 | 80 | 9 | 9 | 9 | 9 |
|
624 |
+
| Spanish | 36 | 50 | 10 | 10 | 10 | 10 |
|
625 |
+
| Italian | 22 | 43 | 5 | 5 | 5 | 5 |
|
626 |
+
| Portuguese | 26 | 16 | 5 | 5 | 5 | 5 |
|
627 |
+
| Polish | 6 | 5 | 2 | 2 | 2 | 2 |
|
628 |
+
|
629 |
+
| # Hours / Gender | Dev | | Test | |
|
630 |
+
|------------------|------|------|------|------|
|
631 |
+
| Gender | M | F | M | F |
|
632 |
+
| English | 7.76 | 7.99 | 7.62 | 7.93 |
|
633 |
+
| German | 7.06 | 7.22 | 7 | 7.29 |
|
634 |
+
| Dutch | 6.44 | 6.32 | 6.72 | 6.04 |
|
635 |
+
| French | 5.13 | 4.94 | 5.04 | 5.02 |
|
636 |
+
| Spanish | 4.91 | 5.08 | 4.78 | 5.23 |
|
637 |
+
| Italian | 2.5 | 2.68 | 2.38 | 2.9 |
|
638 |
+
| Portuguese | 1.84 | 1.81 | 1.83 | 1.9 |
|
639 |
+
| Polish | 1.12 | 0.95 | 1.09 | 1.05 |
|
640 |
+
|
641 |
+
|
642 |
+
|
643 |
+
|
644 |
### Contributions
|
645 |
|
646 |
+
Thanks to [@patrickvonplaten](https://github.com/patrickvonplaten) and [@polinaeterna](https://github.com/polinaeterna) for adding this dataset.
|
|