Update README.md
Browse files
README.md
CHANGED
@@ -37,7 +37,201 @@ configs:
|
|
37 |
data_files:
|
38 |
- split: train
|
39 |
path: male/train-*
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
---
|
41 |
-
# Dataset Card for
|
42 |
|
43 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
data_files:
|
38 |
- split: train
|
39 |
path: male/train-*
|
40 |
+
task_categories:
|
41 |
+
- text-to-speech
|
42 |
+
- text-to-audio
|
43 |
+
language:
|
44 |
+
- es
|
45 |
+
pretty_name: Chilean Spanish Speech
|
46 |
---
|
47 |
+
# Dataset Card for Tamil Speech
|
48 |
|
49 |
+
## Table of Contents
|
50 |
+
- [Dataset Description](#dataset-description)
|
51 |
+
- [Dataset Summary](#dataset-summary)
|
52 |
+
- [Supported Tasks](#supported-tasks)
|
53 |
+
- [How to use](#how-to-use)
|
54 |
+
- [Dataset Structure](#dataset-structure)
|
55 |
+
- [Data Instances](#data-instances)
|
56 |
+
- [Data Fields](#data-fields)
|
57 |
+
- [Data Statistics](#data-statistics)
|
58 |
+
- [Dataset Creation](#dataset-creation)
|
59 |
+
- [Curation Rationale](#curation-rationale)
|
60 |
+
- [Source Data](#source-data)
|
61 |
+
- [Annotations](#annotations)
|
62 |
+
- [Personal and Sensitive Information](#personal-and-sensitive-information)
|
63 |
+
- [Considerations for Using the Data](#considerations-for-using-the-data)
|
64 |
+
- [Social Impact of Dataset](#social-impact-of-dataset)
|
65 |
+
- [Discussion of Biases](#discussion-of-biases)
|
66 |
+
- [Other Known Limitations](#other-known-limitations)
|
67 |
+
- [Additional Information](#additional-information)
|
68 |
+
- [Dataset Curators](#dataset-curators)
|
69 |
+
- [Licensing Information](#licensing-information)
|
70 |
+
- [Citation Information](#citation-information)
|
71 |
+
- [Contributions](#contributions)
|
72 |
+
|
73 |
+
## Dataset Description
|
74 |
+
|
75 |
+
- **Homepage:** [Crowdsourced high-quality Chilean Spanish speech data set.](https://www.openslr.org/71/)
|
76 |
+
- **Repository:** [Google Language Resources and Tools](https://github.com/google/language-resources)
|
77 |
+
- **Paper:** [Crowdsourcing Latin American Spanish for Low-Resource Text-to-Speech](https://aclanthology.org/2020.lrec-1.801/)
|
78 |
+
|
79 |
+
### Dataset Summary
|
80 |
+
|
81 |
+
This dataset consists of 7 hours of transcribed high-quality audio of Chilean Spanish sentences recorded by 31 volunteers. The dataset is intended for speech technologies.
|
82 |
+
|
83 |
+
The data archives were restructured from the original ones from [OpenSLR](http://www.openslr.org/71/) to make it easier to stream.
|
84 |
+
|
85 |
+
|
86 |
+
### Supported Tasks
|
87 |
+
|
88 |
+
- `text-to-speech`, `text-to-audio`: The dataset can be used to train a model for Text-To-Speech (TTS).
|
89 |
+
- `automatic-speech-recognition`, `speaker-identification`: The dataset can also 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).
|
90 |
+
|
91 |
+
|
92 |
+
### How to use
|
93 |
+
|
94 |
+
The `datasets` library allows you to load and pre-process your dataset in pure Python, at scale. The dataset can be downloaded and prepared in one call to your local drive by using the `load_dataset` function.
|
95 |
+
|
96 |
+
For example, to download the female config, simply specify the corresponding language config name (i.e., "female" for female speakers):
|
97 |
+
```python
|
98 |
+
from datasets import load_dataset
|
99 |
+
|
100 |
+
dataset =load_dataset("ylacombe/google-chilean-spanish", "female", split="train")
|
101 |
+
```
|
102 |
+
|
103 |
+
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.
|
104 |
+
```python
|
105 |
+
from datasets import load_dataset
|
106 |
+
|
107 |
+
dataset =load_dataset("ylacombe/google-chilean-spanish", "female", split="train", streaming=True)
|
108 |
+
|
109 |
+
print(next(iter(dataset)))
|
110 |
+
```
|
111 |
+
|
112 |
+
#### *Bonus*
|
113 |
+
You can create a [PyTorch dataloader](https://huggingface.co/docs/datasets/use_with_pytorch) directly with your own datasets (local/streamed).
|
114 |
+
|
115 |
+
**Local:**
|
116 |
+
|
117 |
+
```python
|
118 |
+
from datasets import load_dataset
|
119 |
+
from torch.utils.data.sampler import BatchSampler, RandomSampler
|
120 |
+
|
121 |
+
dataset =load_dataset("ylacombe/google-chilean-spanish", "female", split="train")
|
122 |
+
batch_sampler = BatchSampler(RandomSampler(dataset), batch_size=32, drop_last=False)
|
123 |
+
dataloader = DataLoader(dataset, batch_sampler=batch_sampler)
|
124 |
+
```
|
125 |
+
|
126 |
+
**Streaming:**
|
127 |
+
|
128 |
+
```python
|
129 |
+
from datasets import load_dataset
|
130 |
+
from torch.utils.data import DataLoader
|
131 |
+
|
132 |
+
dataset =load_dataset("ylacombe/google-chilean-spanish", "female", split="train", streaming=True)
|
133 |
+
dataloader = DataLoader(dataset, batch_size=32)
|
134 |
+
```
|
135 |
+
|
136 |
+
To find out more about loading and preparing audio datasets, head over to [hf.co/blog/audio-datasets](https://huggingface.co/blog/audio-datasets).
|
137 |
+
|
138 |
+
## Dataset Structure
|
139 |
+
|
140 |
+
### Data Instances
|
141 |
+
|
142 |
+
A typical data point comprises the path to the audio file called `audio` and its transcription, called `text`. Some additional information about the speaker and the passage which contains the transcription is provided.
|
143 |
+
|
144 |
+
```
|
145 |
+
{'audio': {'path': 'clf_09334_01278378087.wav', 'array': array([-9.15527344e-05, -4.57763672e-04, -4.88281250e-04, ...,
|
146 |
+
1.86157227e-03, 2.10571289e-03, 2.31933594e-03]), 'sampling_rate': 48000}, 'text': 'La vigencia de tu tarjeta es de ocho meses', 'speaker_id': 9334}
|
147 |
+
```
|
148 |
+
|
149 |
+
### Data Fields
|
150 |
+
|
151 |
+
- audio: A dictionary containing the audio filename, 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]`.
|
152 |
+
|
153 |
+
- text: the transcription of the audio file.
|
154 |
+
|
155 |
+
- speaker_id: unique id of the speaker. The same speaker id can be found for multiple data samples.
|
156 |
+
|
157 |
+
### Data Statistics
|
158 |
+
|
159 |
+
| | Total duration (h) | # speakers | # sentences | # total words | # unique words |
|
160 |
+
|--------|--------------------|------------|-------------|---------------|----------------|
|
161 |
+
| Female | 2.84 | 13 | 1738 | 16591 | 3279 |
|
162 |
+
| Male | 4.31 | 18 | 2636 | 25168 | 4171 |
|
163 |
+
|
164 |
+
## Dataset Creation
|
165 |
+
|
166 |
+
### Curation Rationale
|
167 |
+
|
168 |
+
[Needs More Information]
|
169 |
+
|
170 |
+
### Source Data
|
171 |
+
|
172 |
+
#### Initial Data Collection and Normalization
|
173 |
+
|
174 |
+
[Needs More Information]
|
175 |
+
|
176 |
+
#### Who are the source language producers?
|
177 |
+
|
178 |
+
[Needs More Information]
|
179 |
+
|
180 |
+
### Annotations
|
181 |
+
|
182 |
+
#### Annotation process
|
183 |
+
|
184 |
+
[Needs More Information]
|
185 |
+
|
186 |
+
#### Who are the annotators?
|
187 |
+
|
188 |
+
[Needs More Information]
|
189 |
+
|
190 |
+
### Personal and Sensitive Information
|
191 |
+
|
192 |
+
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.
|
193 |
+
|
194 |
+
## Considerations for Using the Data
|
195 |
+
|
196 |
+
### Social Impact of Dataset
|
197 |
+
|
198 |
+
[More Information Needed]
|
199 |
+
|
200 |
+
### Discussion of Biases
|
201 |
+
|
202 |
+
[More Information Needed]
|
203 |
+
|
204 |
+
### Other Known Limitations
|
205 |
+
|
206 |
+
[Needs More Information]
|
207 |
+
|
208 |
+
## Additional Information
|
209 |
+
|
210 |
+
### Dataset Curators
|
211 |
+
|
212 |
+
[Needs More Information]
|
213 |
+
|
214 |
+
### Licensing Information
|
215 |
+
|
216 |
+
License: ([CC BY-SA 4.0 DEED](https://creativecommons.org/licenses/by-sa/4.0/deed.en))
|
217 |
+
|
218 |
+
### Citation Information
|
219 |
+
|
220 |
+
```
|
221 |
+
@inproceedings{guevara-rukoz-etal-2020-crowdsourcing,
|
222 |
+
title = {{Crowdsourcing Latin American Spanish for Low-Resource Text-to-Speech}},
|
223 |
+
author = {Guevara-Rukoz, Adriana and Demirsahin, Isin and He, Fei and Chu, Shan-Hui Cathy and Sarin, Supheakmungkol and Pipatsrisawat, Knot and Gutkin, Alexander and Butryna, Alena and Kjartansson, Oddur},
|
224 |
+
booktitle = {Proceedings of The 12th Language Resources and Evaluation Conference (LREC)},
|
225 |
+
year = {2020},
|
226 |
+
month = may,
|
227 |
+
address = {Marseille, France},
|
228 |
+
publisher = {European Language Resources Association (ELRA)},
|
229 |
+
url = {https://www.aclweb.org/anthology/2020.lrec-1.801},
|
230 |
+
pages = {6504--6513},
|
231 |
+
ISBN = {979-10-95546-34-4},
|
232 |
+
}
|
233 |
+
```
|
234 |
+
|
235 |
+
### Contributions
|
236 |
+
|
237 |
+
Thanks to [@ylacombe](https://github.com/ylacombe) for adding this dataset.
|