Datasets:
Languages:
English
ArXiv:
Tags:
query-by-example-spoken-term-detection
audio-slot-filling
speaker-diarization
automatic-speaker-verification
License:
Commit
•
64cab28
0
Parent(s):
Update files from the datasets library (from 1.10.0)
Browse filesRelease notes: https://github.com/huggingface/datasets/releases/tag/1.10.0
- .gitattributes +27 -0
- README.md +388 -0
- dataset_infos.json +1 -0
- dummy/asr/1.9.0/dummy_data.zip +3 -0
- superb.py +192 -0
.gitattributes
ADDED
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
*.7z filter=lfs diff=lfs merge=lfs -text
|
2 |
+
*.arrow filter=lfs diff=lfs merge=lfs -text
|
3 |
+
*.bin filter=lfs diff=lfs merge=lfs -text
|
4 |
+
*.bin.* filter=lfs diff=lfs merge=lfs -text
|
5 |
+
*.bz2 filter=lfs diff=lfs merge=lfs -text
|
6 |
+
*.ftz filter=lfs diff=lfs merge=lfs -text
|
7 |
+
*.gz filter=lfs diff=lfs merge=lfs -text
|
8 |
+
*.h5 filter=lfs diff=lfs merge=lfs -text
|
9 |
+
*.joblib filter=lfs diff=lfs merge=lfs -text
|
10 |
+
*.lfs.* filter=lfs diff=lfs merge=lfs -text
|
11 |
+
*.model filter=lfs diff=lfs merge=lfs -text
|
12 |
+
*.msgpack filter=lfs diff=lfs merge=lfs -text
|
13 |
+
*.onnx filter=lfs diff=lfs merge=lfs -text
|
14 |
+
*.ot filter=lfs diff=lfs merge=lfs -text
|
15 |
+
*.parquet filter=lfs diff=lfs merge=lfs -text
|
16 |
+
*.pb filter=lfs diff=lfs merge=lfs -text
|
17 |
+
*.pt filter=lfs diff=lfs merge=lfs -text
|
18 |
+
*.pth filter=lfs diff=lfs merge=lfs -text
|
19 |
+
*.rar filter=lfs diff=lfs merge=lfs -text
|
20 |
+
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
21 |
+
*.tar.* filter=lfs diff=lfs merge=lfs -text
|
22 |
+
*.tflite filter=lfs diff=lfs merge=lfs -text
|
23 |
+
*.tgz filter=lfs diff=lfs merge=lfs -text
|
24 |
+
*.xz filter=lfs diff=lfs merge=lfs -text
|
25 |
+
*.zip filter=lfs diff=lfs merge=lfs -text
|
26 |
+
*.zstandard filter=lfs diff=lfs merge=lfs -text
|
27 |
+
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
@@ -0,0 +1,388 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
annotations_creators:
|
3 |
+
- other
|
4 |
+
language_creators:
|
5 |
+
- other
|
6 |
+
languages:
|
7 |
+
- en
|
8 |
+
licenses:
|
9 |
+
- unknown
|
10 |
+
multilinguality:
|
11 |
+
- monolingual
|
12 |
+
pretty_name: SUPERB
|
13 |
+
size_categories:
|
14 |
+
- unknown
|
15 |
+
source_datasets:
|
16 |
+
- original
|
17 |
+
- extended|librispeech_asr
|
18 |
+
task_categories:
|
19 |
+
- speech-processing
|
20 |
+
task_ids:
|
21 |
+
- automatic-speech-recognition
|
22 |
+
- phoneme-recognition
|
23 |
+
- keyword-spotting
|
24 |
+
- query-by-example-spoken-term-detection
|
25 |
+
- speaker-identification
|
26 |
+
- automatic-speaker-verification
|
27 |
+
- speaker-diarization
|
28 |
+
- intent-classification
|
29 |
+
- slot-filling
|
30 |
+
- emotion-recognition
|
31 |
+
---
|
32 |
+
|
33 |
+
# Dataset Card for SUPERB
|
34 |
+
|
35 |
+
## Table of Contents
|
36 |
+
- [Table of Contents](#table-of-contents)
|
37 |
+
- [Dataset Description](#dataset-description)
|
38 |
+
- [Dataset Summary](#dataset-summary)
|
39 |
+
- [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
|
40 |
+
- [Languages](#languages)
|
41 |
+
- [Dataset Structure](#dataset-structure)
|
42 |
+
- [Data Instances](#data-instances)
|
43 |
+
- [Data Fields](#data-fields)
|
44 |
+
- [Data Splits](#data-splits)
|
45 |
+
- [Dataset Creation](#dataset-creation)
|
46 |
+
- [Curation Rationale](#curation-rationale)
|
47 |
+
- [Source Data](#source-data)
|
48 |
+
- [Annotations](#annotations)
|
49 |
+
- [Personal and Sensitive Information](#personal-and-sensitive-information)
|
50 |
+
- [Considerations for Using the Data](#considerations-for-using-the-data)
|
51 |
+
- [Social Impact of Dataset](#social-impact-of-dataset)
|
52 |
+
- [Discussion of Biases](#discussion-of-biases)
|
53 |
+
- [Other Known Limitations](#other-known-limitations)
|
54 |
+
- [Additional Information](#additional-information)
|
55 |
+
- [Dataset Curators](#dataset-curators)
|
56 |
+
- [Licensing Information](#licensing-information)
|
57 |
+
- [Citation Information](#citation-information)
|
58 |
+
- [Contributions](#contributions)
|
59 |
+
|
60 |
+
## Dataset Description
|
61 |
+
|
62 |
+
- **Homepage:** [http://superbbenchmark.org](http://superbbenchmark.org)
|
63 |
+
- **Repository:** [https://github.com/s3prl/s3prl](https://github.com/s3prl/s3prl)
|
64 |
+
- **Paper:** [SUPERB: Speech processing Universal PERformance Benchmark](https://arxiv.org/abs/2105.01051)
|
65 |
+
- **Leaderboard:** [More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
66 |
+
- **Point of Contact:** [Lewis Tunstall](mailto:lewis@huggingface.co) and [Albert Villanova](mailto:albert@huggingface.co)
|
67 |
+
|
68 |
+
### Dataset Summary
|
69 |
+
|
70 |
+
SUPERB is a leaderboard to benchmark the performance of a shared model across a wide range of speech processing tasks with minimal architecture changes and labeled data.
|
71 |
+
|
72 |
+
### Supported Tasks and Leaderboards
|
73 |
+
|
74 |
+
The SUPERB leaderboard can be found here **ADD LINK WHEN LIVE** and consists of the following tasks:
|
75 |
+
|
76 |
+
#### pr
|
77 |
+
|
78 |
+
Phoneme Recognition (PR) transcribes an utterance into the smallest content units. This task includes alignment modeling to avoid potentially inaccurate forced alignment. [LibriSpeech](https://huggingface.co/datasets/librispeech_asr) train-clean-100/dev-clean/test-clean subsets are adopted in SUPERB for training/validation/testing. Phoneme transcriptions are obtained from the LibriSpeech official g2p-model-5 and the conversion script in Kaldi librispeech s5 recipe. The evaluation metric is phone error rate (PER).
|
79 |
+
|
80 |
+
#### asr
|
81 |
+
|
82 |
+
Automatic Speech Recognition (ASR) transcribes utterances into words. While PR analyzes the improvement in modeling phonetics, ASR reflects the significance of the improvement in a real-world scenario. [LibriSpeech](https://huggingface.co/datasets/librispeech_asr) train-clean-100/devclean/test-clean subsets are used for training/validation/testing. The evaluation metric is word error rate (WER).
|
83 |
+
|
84 |
+
#### ks
|
85 |
+
|
86 |
+
Keyword Spotting (KS) detects preregistered keywords by classifying utterances into a predefined set of words. The task is usually performed on-device for the fast response time. Thus, accuracy, model size, and inference time are all crucial. SUPERB uses the widely used [Speech Commands dataset v1.0](https://www.tensorflow.org/datasets/catalog/speech_commands) for the task. The dataset consists of ten classes of keywords, a class for silence, and an unknown class to include the false positive. The evaluation metric is accuracy (ACC)
|
87 |
+
|
88 |
+
#### qbe
|
89 |
+
|
90 |
+
Query by Example Spoken Term Detection (QbE) detects a spoken term (query) in an audio database (documents) by binary discriminating a given pair of query and document into a match or not. The English subset in [QUESST 2014 challenge](https://github.com/s3prl/s3prl/tree/master/downstream#qbe-query-by-example-spoken-term-detection) is adopted since we focus on investigating English as the first step. The evaluation metric is maximum term weighted value (MTWV) which balances misses and false alarms.
|
91 |
+
|
92 |
+
#### ic
|
93 |
+
|
94 |
+
Intent Classification (IC) classifies utterances into predefined classes to determine the intent of speakers. SUPERB uses the [Fluent Speech Commands dataset](https://github.com/s3prl/s3prl/tree/master/downstream#ic-intent-classification---fluent-speech-commands), where each utterance is tagged with three intent labels: action, object, and location. The evaluation metric is accuracy (ACC).
|
95 |
+
|
96 |
+
#### sf
|
97 |
+
|
98 |
+
Slot Filling (SF) predicts a sequence of semantic slot-types from an utterance, like a slot-type FromLocation for a spoken word Taipei, which is known as a slot-value. Both slot-types and slot-values are essential for an SLU system to function. The evaluation metrics thus include slot-type F1 score and slotvalue CER. [Audio SNIPS](https://github.com/s3prl/s3prl/tree/master/downstream#sf-end-to-end-slot-filling) is adopted, which synthesized multi-speaker utterances for SNIPS. Following the standard split in SNIPS, US-accent speakers are further selected for training, and others are for validation/testing.
|
99 |
+
|
100 |
+
#### si
|
101 |
+
Speaker Identification (SI) classifies each utterance for its speaker identity as a multi-class classification, where speakers are in the same predefined set for both training and testing. The widely used [VoxCeleb1 dataset](https://www.robots.ox.ac.uk/~vgg/data/voxceleb/vox1.html) is adopted, and the evaluation metric is accuracy (ACC).
|
102 |
+
|
103 |
+
#### asv
|
104 |
+
|
105 |
+
Automatic Speaker Verification (ASV) verifies whether the speakers of a pair of utterances match as a binary classification, and speakers in the testing set may not appear in the training set. Thus, ASV is more challenging than SID. VoxCeleb1 is used without VoxCeleb2 training data and noise augmentation. The evaluation metric is equal error rate (EER).
|
106 |
+
|
107 |
+
#### sd
|
108 |
+
|
109 |
+
Speaker Diarization (SD) predicts who is speaking when for each timestamp, and multiple speakers can speak simultaneously. The model has to encode rich speaker characteristics for each frame and should be able to represent mixtures of signals. [LibriMix](https://github.com/s3prl/s3prl/tree/master/downstream#sd-speaker-diarization) is adopted where LibriSpeech train-clean-100/dev-clean/test-clean are used to generate mixtures for training/validation/testing. We focus on the two-speaker scenario as the first step. The time-coded speaker labels were generated using alignments from Kaldi LibriSpeech ASR model. The evaluation metric is diarization error rate (DER).
|
110 |
+
|
111 |
+
#### er
|
112 |
+
|
113 |
+
Emotion Recognition (ER) predicts an emotion class for each utterance. The most widely used ER dataset [IEMOCAP](https://github.com/s3prl/s3prl/tree/master/downstream#er-emotion-recognition) is adopted, and we follow the conventional evaluation protocol: we drop the unbalance emotion classes to leave the final four classes with a similar amount of data points and cross-validates on five folds of the standard splits. The evaluation metric is accuracy (ACC).
|
114 |
+
|
115 |
+
### Languages
|
116 |
+
|
117 |
+
The language data in SUPERB is in English (BCP-47 `en`)
|
118 |
+
|
119 |
+
|
120 |
+
## Dataset Structure
|
121 |
+
|
122 |
+
### Data Instances
|
123 |
+
|
124 |
+
#### pr
|
125 |
+
|
126 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
127 |
+
|
128 |
+
|
129 |
+
#### asr
|
130 |
+
|
131 |
+
An example from each split looks like:
|
132 |
+
|
133 |
+
```json
|
134 |
+
{'chapter_id': 1240,
|
135 |
+
'file': 'path/to/file.flac',
|
136 |
+
'id': '103-1240-0000',
|
137 |
+
'speaker_id': 103,
|
138 |
+
'text': 'CHAPTER ONE MISSUS RACHEL LYNDE IS SURPRISED MISSUS RACHEL LYNDE '
|
139 |
+
'LIVED JUST WHERE THE AVONLEA MAIN ROAD DIPPED DOWN INTO A LITTLE '
|
140 |
+
'HOLLOW FRINGED WITH ALDERS AND LADIES EARDROPS AND TRAVERSED BY A '
|
141 |
+
'BROOK'}
|
142 |
+
```
|
143 |
+
|
144 |
+
#### ks
|
145 |
+
|
146 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
147 |
+
|
148 |
+
|
149 |
+
#### qbe
|
150 |
+
|
151 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
152 |
+
|
153 |
+
|
154 |
+
#### ic
|
155 |
+
|
156 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
157 |
+
|
158 |
+
|
159 |
+
#### sf
|
160 |
+
|
161 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
162 |
+
|
163 |
+
|
164 |
+
#### si
|
165 |
+
|
166 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
167 |
+
|
168 |
+
|
169 |
+
#### asv
|
170 |
+
|
171 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
172 |
+
|
173 |
+
|
174 |
+
#### sd
|
175 |
+
|
176 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
177 |
+
|
178 |
+
|
179 |
+
#### er
|
180 |
+
|
181 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
182 |
+
|
183 |
+
|
184 |
+
|
185 |
+
|
186 |
+
### Data Fields
|
187 |
+
|
188 |
+
#### pr
|
189 |
+
|
190 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
191 |
+
|
192 |
+
|
193 |
+
#### asr
|
194 |
+
|
195 |
+
- `file`: a `string` feature.
|
196 |
+
- `text`: a `string` feature.
|
197 |
+
- `speaker_id`: a `int64` feature
|
198 |
+
- `chapter_id`: a `int64` feature
|
199 |
+
- `id`: a `string` feature
|
200 |
+
|
201 |
+
#### ks
|
202 |
+
|
203 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
204 |
+
|
205 |
+
|
206 |
+
#### qbe
|
207 |
+
|
208 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
209 |
+
|
210 |
+
|
211 |
+
#### ic
|
212 |
+
|
213 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
214 |
+
|
215 |
+
|
216 |
+
#### sf
|
217 |
+
|
218 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
219 |
+
|
220 |
+
|
221 |
+
#### si
|
222 |
+
|
223 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
224 |
+
|
225 |
+
|
226 |
+
#### asv
|
227 |
+
|
228 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
229 |
+
|
230 |
+
|
231 |
+
#### sd
|
232 |
+
|
233 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
234 |
+
|
235 |
+
|
236 |
+
#### er
|
237 |
+
|
238 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
239 |
+
|
240 |
+
### Data Splits
|
241 |
+
|
242 |
+
#### pr
|
243 |
+
|
244 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
245 |
+
|
246 |
+
|
247 |
+
#### asr
|
248 |
+
|
249 |
+
| | train | validation | test |
|
250 |
+
|-----|------:|-----------:|-----:|
|
251 |
+
| asr | 28539 | 2703 | 2620 |
|
252 |
+
|
253 |
+
#### ks
|
254 |
+
|
255 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
256 |
+
|
257 |
+
|
258 |
+
#### qbe
|
259 |
+
|
260 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
261 |
+
|
262 |
+
|
263 |
+
#### ic
|
264 |
+
|
265 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
266 |
+
|
267 |
+
|
268 |
+
#### sf
|
269 |
+
|
270 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
271 |
+
|
272 |
+
|
273 |
+
#### si
|
274 |
+
|
275 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
276 |
+
|
277 |
+
|
278 |
+
#### asv
|
279 |
+
|
280 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
281 |
+
|
282 |
+
|
283 |
+
#### sd
|
284 |
+
|
285 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
286 |
+
|
287 |
+
|
288 |
+
#### er
|
289 |
+
|
290 |
+
[More Information Needed](https://github.com/huggingface/datasets/blob/master/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
|
291 |
+
|
292 |
+
## Dataset Creation
|
293 |
+
|
294 |
+
### Curation Rationale
|
295 |
+
|
296 |
+
[More Information Needed]
|
297 |
+
|
298 |
+
### Source Data
|
299 |
+
|
300 |
+
#### Initial Data Collection and Normalization
|
301 |
+
|
302 |
+
[More Information Needed]
|
303 |
+
|
304 |
+
#### Who are the source language producers?
|
305 |
+
|
306 |
+
[More Information Needed]
|
307 |
+
|
308 |
+
### Annotations
|
309 |
+
|
310 |
+
#### Annotation process
|
311 |
+
|
312 |
+
[More Information Needed]
|
313 |
+
|
314 |
+
#### Who are the annotators?
|
315 |
+
|
316 |
+
[More Information Needed]
|
317 |
+
|
318 |
+
### Personal and Sensitive Information
|
319 |
+
|
320 |
+
[More Information Needed]
|
321 |
+
|
322 |
+
## Considerations for Using the Data
|
323 |
+
|
324 |
+
### Social Impact of Dataset
|
325 |
+
|
326 |
+
[More Information Needed]
|
327 |
+
|
328 |
+
### Discussion of Biases
|
329 |
+
|
330 |
+
[More Information Needed]
|
331 |
+
|
332 |
+
### Other Known Limitations
|
333 |
+
|
334 |
+
[More Information Needed]
|
335 |
+
|
336 |
+
## Additional Information
|
337 |
+
|
338 |
+
### Dataset Curators
|
339 |
+
|
340 |
+
[More Information Needed]
|
341 |
+
|
342 |
+
### Licensing Information
|
343 |
+
|
344 |
+
[More Information Needed]
|
345 |
+
|
346 |
+
### Citation Information
|
347 |
+
|
348 |
+
```
|
349 |
+
@article{DBLP:journals/corr/abs-2105-01051,
|
350 |
+
author = {Shu{-}Wen Yang and
|
351 |
+
Po{-}Han Chi and
|
352 |
+
Yung{-}Sung Chuang and
|
353 |
+
Cheng{-}I Jeff Lai and
|
354 |
+
Kushal Lakhotia and
|
355 |
+
Yist Y. Lin and
|
356 |
+
Andy T. Liu and
|
357 |
+
Jiatong Shi and
|
358 |
+
Xuankai Chang and
|
359 |
+
Guan{-}Ting Lin and
|
360 |
+
Tzu{-}Hsien Huang and
|
361 |
+
Wei{-}Cheng Tseng and
|
362 |
+
Ko{-}tik Lee and
|
363 |
+
Da{-}Rong Liu and
|
364 |
+
Zili Huang and
|
365 |
+
Shuyan Dong and
|
366 |
+
Shang{-}Wen Li and
|
367 |
+
Shinji Watanabe and
|
368 |
+
Abdelrahman Mohamed and
|
369 |
+
Hung{-}yi Lee},
|
370 |
+
title = {{SUPERB:} Speech processing Universal PERformance Benchmark},
|
371 |
+
journal = {CoRR},
|
372 |
+
volume = {abs/2105.01051},
|
373 |
+
year = {2021},
|
374 |
+
url = {https://arxiv.org/abs/2105.01051},
|
375 |
+
archivePrefix = {arXiv},
|
376 |
+
eprint = {2105.01051},
|
377 |
+
timestamp = {Thu, 01 Jul 2021 13:30:22 +0200},
|
378 |
+
biburl = {https://dblp.org/rec/journals/corr/abs-2105-01051.bib},
|
379 |
+
bibsource = {dblp computer science bibliography, https://dblp.org}
|
380 |
+
}
|
381 |
+
|
382 |
+
Note that each SUPERB dataset has its own citation. Please see the source to see
|
383 |
+
the correct citation for each contained dataset.
|
384 |
+
```
|
385 |
+
|
386 |
+
### Contributions
|
387 |
+
|
388 |
+
Thanks to [@lewtun](https://github.com/lewtun) and [@albertvillanova](https://github.com/albertvillanova) for adding this dataset.
|
dataset_infos.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"asr": {"description": "Self-supervised learning (SSL) has proven vital for advancing research in\nnatural language processing (NLP) and computer vision (CV). The paradigm\npretrains a shared model on large volumes of unlabeled data and achieves\nstate-of-the-art (SOTA) for various tasks with minimal adaptation. However, the\nspeech processing community lacks a similar setup to systematically explore the\nparadigm. To bridge this gap, we introduce Speech processing Universal\nPERformance Benchmark (SUPERB). SUPERB is a leaderboard to benchmark the\nperformance of a shared model across a wide range of speech processing tasks\nwith minimal architecture changes and labeled data. Among multiple usages of the\nshared model, we especially focus on extracting the representation learned from\nSSL due to its preferable re-usability. We present a simple framework to solve\nSUPERB tasks by learning task-specialized lightweight prediction heads on top of\nthe frozen shared model. Our results demonstrate that the framework is promising\nas SSL representations show competitive generalizability and accessibility\nacross SUPERB tasks. We release SUPERB as a challenge with a leaderboard and a\nbenchmark toolkit to fuel the research in representation learning and general\nspeech processing.\n\nNote that in order to limit the required storage for preparing this dataset, the\naudio is stored in the .flac format and is not converted to a float32 array. To\nconvert, the audio file to a float32 array, please make use of the `.map()`\nfunction as follows:\n\n\n```python\nimport soundfile as sf\n\ndef map_to_array(batch):\n speech_array, _ = sf.read(batch[\"file\"])\n batch[\"speech\"] = speech_array\n return batch\n\ndataset = dataset.map(map_to_array, remove_columns=[\"file\"])\n```\n", "citation": "@article{DBLP:journals/corr/abs-2105-01051,\n author = {Shu{-}Wen Yang and\n Po{-}Han Chi and\n Yung{-}Sung Chuang and\n Cheng{-}I Jeff Lai and\n Kushal Lakhotia and\n Yist Y. Lin and\n Andy T. Liu and\n Jiatong Shi and\n Xuankai Chang and\n Guan{-}Ting Lin and\n Tzu{-}Hsien Huang and\n Wei{-}Cheng Tseng and\n Ko{-}tik Lee and\n Da{-}Rong Liu and\n Zili Huang and\n Shuyan Dong and\n Shang{-}Wen Li and\n Shinji Watanabe and\n Abdelrahman Mohamed and\n Hung{-}yi Lee},\n title = {{SUPERB:} Speech processing Universal PERformance Benchmark},\n journal = {CoRR},\n volume = {abs/2105.01051},\n year = {2021},\n url = {https://arxiv.org/abs/2105.01051},\n archivePrefix = {arXiv},\n eprint = {2105.01051},\n timestamp = {Thu, 01 Jul 2021 13:30:22 +0200},\n biburl = {https://dblp.org/rec/journals/corr/abs-2105-01051.bib},\n bibsource = {dblp computer science bibliography, https://dblp.org}\n}\n", "homepage": "http://www.openslr.org/12", "license": "", "features": {"file": {"dtype": "string", "id": null, "_type": "Value"}, "text": {"dtype": "string", "id": null, "_type": "Value"}, "speaker_id": {"dtype": "int64", "id": null, "_type": "Value"}, "chapter_id": {"dtype": "int64", "id": null, "_type": "Value"}, "id": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": {"input": "file", "output": "text"}, "task_templates": [{"task": "automatic-speech-recognition", "audio_file_path_column": "file", "transcription_column": "text"}], "builder_name": "superb", "config_name": "asr", "version": {"version_str": "1.9.0", "description": "", "major": 1, "minor": 9, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 11823891, "num_examples": 28539, "dataset_name": "superb"}, "validation": {"name": "validation", "num_bytes": 894510, "num_examples": 2703, "dataset_name": "superb"}, "test": {"name": "test", "num_bytes": 868614, "num_examples": 2620, "dataset_name": "superb"}}, "download_checksums": {"http://www.openslr.org/resources/12/dev-clean.tar.gz": {"num_bytes": 337926286, "checksum": "76f87d090650617fca0cac8f88b9416e0ebf80350acb97b343a85fa903728ab3"}, "http://www.openslr.org/resources/12/test-clean.tar.gz": {"num_bytes": 346663984, "checksum": "39fde525e59672dc6d1551919b1478f724438a95aa55f874b576be21967e6c23"}, "http://www.openslr.org/resources/12/train-clean-100.tar.gz": {"num_bytes": 6387309499, "checksum": "d4ddd1d5a6ab303066f14971d768ee43278a5f2a0aa43dc716b0e64ecbbbf6e2"}}, "download_size": 7071899769, "post_processing_size": null, "dataset_size": 13587015, "size_in_bytes": 7085486784}}
|
dummy/asr/1.9.0/dummy_data.zip
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3ff8f5cf44cc18aa16659d42a5ef7ebd754ef594130762f9e6d48909292154b8
|
3 |
+
size 805749
|
superb.py
ADDED
@@ -0,0 +1,192 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# coding=utf-8
|
2 |
+
# Copyright 2021 The TensorFlow Datasets Authors and the HuggingFace Datasets Authors.
|
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 |
+
|
16 |
+
# Lint as: python3
|
17 |
+
"""SUPERB: Speech processing Universal PERformance Benchmark."""
|
18 |
+
|
19 |
+
|
20 |
+
import glob
|
21 |
+
import os
|
22 |
+
import textwrap
|
23 |
+
|
24 |
+
import datasets
|
25 |
+
from datasets.tasks import AutomaticSpeechRecognition
|
26 |
+
|
27 |
+
|
28 |
+
_CITATION = """\
|
29 |
+
@article{DBLP:journals/corr/abs-2105-01051,
|
30 |
+
author = {Shu{-}Wen Yang and
|
31 |
+
Po{-}Han Chi and
|
32 |
+
Yung{-}Sung Chuang and
|
33 |
+
Cheng{-}I Jeff Lai and
|
34 |
+
Kushal Lakhotia and
|
35 |
+
Yist Y. Lin and
|
36 |
+
Andy T. Liu and
|
37 |
+
Jiatong Shi and
|
38 |
+
Xuankai Chang and
|
39 |
+
Guan{-}Ting Lin and
|
40 |
+
Tzu{-}Hsien Huang and
|
41 |
+
Wei{-}Cheng Tseng and
|
42 |
+
Ko{-}tik Lee and
|
43 |
+
Da{-}Rong Liu and
|
44 |
+
Zili Huang and
|
45 |
+
Shuyan Dong and
|
46 |
+
Shang{-}Wen Li and
|
47 |
+
Shinji Watanabe and
|
48 |
+
Abdelrahman Mohamed and
|
49 |
+
Hung{-}yi Lee},
|
50 |
+
title = {{SUPERB:} Speech processing Universal PERformance Benchmark},
|
51 |
+
journal = {CoRR},
|
52 |
+
volume = {abs/2105.01051},
|
53 |
+
year = {2021},
|
54 |
+
url = {https://arxiv.org/abs/2105.01051},
|
55 |
+
archivePrefix = {arXiv},
|
56 |
+
eprint = {2105.01051},
|
57 |
+
timestamp = {Thu, 01 Jul 2021 13:30:22 +0200},
|
58 |
+
biburl = {https://dblp.org/rec/journals/corr/abs-2105-01051.bib},
|
59 |
+
bibsource = {dblp computer science bibliography, https://dblp.org}
|
60 |
+
}
|
61 |
+
"""
|
62 |
+
|
63 |
+
_DESCRIPTION = """\
|
64 |
+
Self-supervised learning (SSL) has proven vital for advancing research in
|
65 |
+
natural language processing (NLP) and computer vision (CV). The paradigm
|
66 |
+
pretrains a shared model on large volumes of unlabeled data and achieves
|
67 |
+
state-of-the-art (SOTA) for various tasks with minimal adaptation. However, the
|
68 |
+
speech processing community lacks a similar setup to systematically explore the
|
69 |
+
paradigm. To bridge this gap, we introduce Speech processing Universal
|
70 |
+
PERformance Benchmark (SUPERB). SUPERB is a leaderboard to benchmark the
|
71 |
+
performance of a shared model across a wide range of speech processing tasks
|
72 |
+
with minimal architecture changes and labeled data. Among multiple usages of the
|
73 |
+
shared model, we especially focus on extracting the representation learned from
|
74 |
+
SSL due to its preferable re-usability. We present a simple framework to solve
|
75 |
+
SUPERB tasks by learning task-specialized lightweight prediction heads on top of
|
76 |
+
the frozen shared model. Our results demonstrate that the framework is promising
|
77 |
+
as SSL representations show competitive generalizability and accessibility
|
78 |
+
across SUPERB tasks. We release SUPERB as a challenge with a leaderboard and a
|
79 |
+
benchmark toolkit to fuel the research in representation learning and general
|
80 |
+
speech processing.
|
81 |
+
|
82 |
+
Note that in order to limit the required storage for preparing this dataset, the
|
83 |
+
audio is stored in the .flac format and is not converted to a float32 array. To
|
84 |
+
convert, the audio file to a float32 array, please make use of the `.map()`
|
85 |
+
function as follows:
|
86 |
+
|
87 |
+
|
88 |
+
```python
|
89 |
+
import soundfile as sf
|
90 |
+
|
91 |
+
def map_to_array(batch):
|
92 |
+
speech_array, _ = sf.read(batch["file"])
|
93 |
+
batch["speech"] = speech_array
|
94 |
+
return batch
|
95 |
+
|
96 |
+
dataset = dataset.map(map_to_array, remove_columns=["file"])
|
97 |
+
```
|
98 |
+
"""
|
99 |
+
|
100 |
+
|
101 |
+
class SuperbConfig(datasets.BuilderConfig):
|
102 |
+
"""BuilderConfig for Superb."""
|
103 |
+
|
104 |
+
def __init__(
|
105 |
+
self,
|
106 |
+
data_url,
|
107 |
+
url,
|
108 |
+
task_templates=None,
|
109 |
+
**kwargs,
|
110 |
+
):
|
111 |
+
super(SuperbConfig, self).__init__(version=datasets.Version("1.9.0", ""), **kwargs)
|
112 |
+
self.data_url = data_url
|
113 |
+
self.url = url
|
114 |
+
self.task_templates = task_templates
|
115 |
+
|
116 |
+
|
117 |
+
class Superb(datasets.GeneratorBasedBuilder):
|
118 |
+
"""Superb dataset."""
|
119 |
+
|
120 |
+
BUILDER_CONFIGS = [
|
121 |
+
SuperbConfig(
|
122 |
+
name="asr",
|
123 |
+
description=textwrap.dedent(
|
124 |
+
"""\
|
125 |
+
ASR transcribes utterances into words. While PR analyzes the
|
126 |
+
improvement in modeling phonetics, ASR reflects the significance of
|
127 |
+
the improvement in a real-world scenario. LibriSpeech
|
128 |
+
train-clean-100/dev-clean/test-clean subsets are used for
|
129 |
+
training/validation/testing. The evaluation metric is word error
|
130 |
+
rate (WER)."""
|
131 |
+
),
|
132 |
+
url="http://www.openslr.org/12",
|
133 |
+
data_url="http://www.openslr.org/resources/12/",
|
134 |
+
task_templates=[AutomaticSpeechRecognition(audio_file_path_column="file", transcription_column="text")],
|
135 |
+
)
|
136 |
+
]
|
137 |
+
|
138 |
+
def _info(self):
|
139 |
+
return datasets.DatasetInfo(
|
140 |
+
description=_DESCRIPTION,
|
141 |
+
features=datasets.Features(
|
142 |
+
{
|
143 |
+
"file": datasets.Value("string"),
|
144 |
+
"text": datasets.Value("string"),
|
145 |
+
"speaker_id": datasets.Value("int64"),
|
146 |
+
"chapter_id": datasets.Value("int64"),
|
147 |
+
"id": datasets.Value("string"),
|
148 |
+
}
|
149 |
+
),
|
150 |
+
supervised_keys=("file", "text"),
|
151 |
+
homepage=self.config.url,
|
152 |
+
citation=_CITATION,
|
153 |
+
task_templates=self.config.task_templates,
|
154 |
+
)
|
155 |
+
|
156 |
+
def _split_generators(self, dl_manager):
|
157 |
+
if self.config.name == "asr":
|
158 |
+
_DL_URLS = {
|
159 |
+
"dev": self.config.data_url + "dev-clean.tar.gz",
|
160 |
+
"test": self.config.data_url + "test-clean.tar.gz",
|
161 |
+
"train": self.config.data_url + "train-clean-100.tar.gz",
|
162 |
+
}
|
163 |
+
archive_path = dl_manager.download_and_extract(_DL_URLS)
|
164 |
+
|
165 |
+
return [
|
166 |
+
datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"archive_path": archive_path["train"]}),
|
167 |
+
datasets.SplitGenerator(
|
168 |
+
name=datasets.Split.VALIDATION, gen_kwargs={"archive_path": archive_path["dev"]}
|
169 |
+
),
|
170 |
+
datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={"archive_path": archive_path["test"]}),
|
171 |
+
]
|
172 |
+
|
173 |
+
def _generate_examples(self, archive_path):
|
174 |
+
"""Generate examples."""
|
175 |
+
transcripts_glob = os.path.join(archive_path, "LibriSpeech", "*/*/*/*.txt")
|
176 |
+
key = 0
|
177 |
+
for transcript_path in sorted(glob.glob(transcripts_glob)):
|
178 |
+
transcript_dir_path = os.path.dirname(transcript_path)
|
179 |
+
with open(transcript_path, "r", encoding="utf-8") as f:
|
180 |
+
for line in f:
|
181 |
+
line = line.strip()
|
182 |
+
id_, transcript = line.split(" ", 1)
|
183 |
+
audio_file = f"{id_}.flac"
|
184 |
+
speaker_id, chapter_id = [int(el) for el in id_.split("-")[:2]]
|
185 |
+
yield key, {
|
186 |
+
"id": id_,
|
187 |
+
"speaker_id": speaker_id,
|
188 |
+
"chapter_id": chapter_id,
|
189 |
+
"file": os.path.join(transcript_dir_path, audio_file),
|
190 |
+
"text": transcript,
|
191 |
+
}
|
192 |
+
key += 1
|