Коллекция образцов речи из различных аудиоисточников. Набор данных содержит короткие аудиоклипы на русском языке.
Вероятно, это самый крупный на данный момент общедоступный набор данных для преобразования речи в текст на русском языке:
- около 16 млн речевых фрагментов;
- около 20 000 часов;
- объем 2,3 ТБ (несжатые файлы
.wav
вint16
) или 356 ГБ (в формате.opus
); - все файлы, кроме наборов данных для проверки, преобразованы в формат OPUS;
основное назначение набора данных — обучение моделей преобразования речи в текст.
Состав набора данныхРазмер набора данных приводится для файлов в формате .wav
.
Dataset | Высказывания | Часы | ГБ | Секунд/символов | Комментировать | Заметка | Качество/шум |
---|---|---|---|---|---|---|---|
radio_v4 (*) | 7 603 192 | 10 430 | 1195 | 5 с / 68 | Переключатель | Выровнять | 95 % / четкое |
public_speech (*) | 1 700 060 | 2709 | 301 | 6 с / 79 | Публичные выступления | Выровнять | 95 % / четкое |
audiobook_2 | 1 149 404 | 1511 | 162 | 5 с / 56 | Книги | Выровнять | 95 % / четкое |
radio_2 | 651 645 | 1439 | 154 | 8 с / 110 | Переключатель | Выровнять | 95 % / четкое |
public_youtube1120 | 1 410 979 | 1104 | 237 | 3 с / 34 | YouTube | Субтитры | 95 % / четкое |
public_youtube700 | 759 483 | 701 | 75 | 3 с / 43 | YouTube | Субтитры | 95 % / четкое |
tts_russian_addresses | 1 741 838 | 754 | 81 | 2 с / 20 | Адреса | 4 голоса для преобразования текста в речь | 100 % / четкое |
asr_public_phone_calls_2 | 603 797 | 601 | 66 | 4 с / 37 | Телефонные звонки | ASR | 70 % / с помехами |
public_youtube1120_hq | 369 245 | 291 | 31 | 3 с / 37 | Главный офис YouTube | Субтитры | 95 % / четкое |
asr_public_phone_calls_1 | 233 868 | 211 | 23 | 3 с / 29 | Телефонные звонки | ASR | 70 % / с помехами |
radio_v4_add (*) | 92 679 | 157 | 18 | 6 с / 80 | Переключатель | Выровнять | 95 % / четкое |
asr_public_stories_2 | 78 186 | 78 | 9 | 4 с / 43 | Книги | ASR | 80 % / четкое |
asr_public_stories_1 | 46 142 | 38 | 4 | 3 с / 30 | Книги | ASR | 80 % / четкое |
public_series_1 | 20 243 | 17 | 2 | 3 с / 38 | YouTube | Субтитры | 95 % / четкое |
asr_calls_2_val | 12 950 | 7,7 | 2 | 2 с / 34 | Телефонные звонки | Заметки вручную | 99 % / четкое |
public_lecture_1 | 6 803 | 6 | 1 | 3 с / 47 | Лекции | Субтитры | 95 % / четкое |
buriy_audiobooks_2_val | 7 850 | 4,9 | 1 | 2 с / 31 | Книги | Заметки вручную | 99 % / четкое |
public_youtube700_val | 7 311 | 4,5 | 1 | 2 с / 35 | YouTube | Заметки вручную | 99 % / четкое |
(*) Для TXT-файлов предоставляется только пример данных.
Методология создания заметокНабор данных собран из открытых источников. Все длинные аудиофрагменты разбиваются на блоки с учетом интенсивности голоса и выравнивания. Некоторые типы аудиофрагментов снабжаются автоматическими заметками и проверяются с применением статистических и эвристических методов.
Объем и частота обновления данныхОбщий объем набора данных составляет 350 ГБ. Общий размер набора данных с общедоступными метками составляет 130 ГБ.
Обновление набора данных для обеспечения обратной совместимости не планируется. Сравнительные тексты и файлы исключений вы можете найти в исходном репозитории.
В будущем будут добавляться новые предметные области и языки.
Нормализация звука
Все файлы нормализуются для более простой и быстрой корректировки и обработки при воспроизведении, как описано далее.
- Преобразование в моно при необходимости.
- Частота выборки изменяется до 16 кГц при необходимости.
- Сохранение в формате 16-разрядных целых чисел.
- Преобразование в формат OPUS.
Применяется метод базы данных на диске.
Каждый аудиофайл (WAV и двоичные файлы) хэширован. На основе этих хэшей создана иерархия папок, чтобы обеспечить оптимизацию операций файловой системы.
target_format = 'wav'
wavb = wav.tobytes()
f_hash = hashlib.sha1(wavb).hexdigest()
store_path = Path(root_folder,
f_hash[0],
f_hash[1:3],
f_hash[3:15] + '.' + target_format)
Набор данных предоставляется в двух форматах.
- Архивы доступны в хранилище BLOB-объектов Azure и (или) по прямым ссылкам.
- Исходные файлы доступны в хранилище BLOB-объектов Azure.
Все данные хранятся по адресу https://azureopendatastorage.blob.core.windows.net/openstt/
Структура папок
└── ru_open_stt_opus <= archived folders
│ │
│ ├── archives
│ │ ├── asr_calls_2_val.tar.gz <= tar.gz archives with opus and wav files
│ │ │ ... <= see the below table for enumeration
│ │ └── tts_russian_addresses_rhvoice_4voices.tar.gz
│ │
│ └── manifests
│ ├── asr_calls_2_val.csv <= csv files with wav_path, text_path, duration (see notebooks)
│ │ ...
│ └── tts_russian_addresses_rhvoice_4voices.csv
│
└── ru_open_stt_opus_unpacked <= a separate folder for each uploaded domain
├── public_youtube1120
│ ├── 0 <= see "On disk DB methodology" for details
│ ├── 1
│ │ ├── 00
│ │ │ ...
│ │ └── ff
│ │ ├── *.opus <= actual files
│ │ └── *.txt
│ │ ...
│ └── f
│
├── public_youtube1120_hq
├── public_youtube700_val
├── asr_calls_2_val
├── radio_2
├── private_buriy_audiobooks_2
├── asr_public_phone_calls_2
├── asr_public_stories_2
├── asr_public_stories_1
├── public_lecture_1
├── asr_public_phone_calls_1
├── public_series_1
└── public_youtube700
Dataset | ГБ, WAV | ГБ, архив | Архив | Источник | манифеста |
---|---|---|---|---|---|
Обучение | |||||
Примеры радиотрансляций и публичных выступлений | - | 11,4 | opus+txt | - | manifest |
audiobook_2 | 162 | 25,8 | opus+txt | Интернет + выравнивание | manifest |
radio_2 | 154 | 24,6 | opus+txt | Переключатель | manifest |
public_youtube1120 | 237 | 19,0 | opus+txt | видео YouTube; | manifest |
asr_public_phone_calls_2 | 66 | 9,4 | opus+txt | Интернет + ASR | manifest |
public_youtube1120_hq | 31 | 4.9 | opus+txt | видео YouTube; | manifest |
asr_public_stories_2 | 9 | 1.4 | opus+txt | Интернет + выравнивание | manifest |
tts_russian_addresses_rhvoice_4voices | 80,9 | 12,9 | opus+txt | TTS | manifest |
public_youtube700 | 75,0 | 12,2 | opus+txt | видео YouTube; | manifest |
asr_public_phone_calls_1 | 22,7 | 3.2 | opus+txt | Интернет + ASR | manifest |
asr_public_stories_1 | 4.1 | 0,7 | opus+txt | Общедоступные истории | manifest |
public_series_1 | 1,9 | 0,3 | opus+txt | Общедоступные серии | manifest |
public_lecture_1 | 0,7 | 0,1 | opus+txt | Интернет + руководства | manifest |
Val | |||||
asr_calls_2_val | 2 | 0,8 | wav+txt | Интернет | manifest |
buriy_audiobooks_2_val | 1 | 0,5 | wav+txt | Книги + руководства | manifest |
public_youtube700_val | 2 | 0.13 | wav+txt | Видео на YouTube + руководство | manifest |
Напрямую
Дополнительные сведения — https://github.com/snakers4/open_stt#download-instructions
Через подключение хранилища BLOB-объектов
См. записную книжку, расположенную на вкладке “Доступ к данным”
КонтактыЕсли вам требуется помощь или у вас возникли вопросы, вы можете обратиться к авторам данных по адресу aveysov@gmail.com.
ЛицензияЭта лицензия разрешает распространять, рекомбинировать, адаптировать и использовать материал в качестве основы на любых носителях и в любом формате только в некоммерческих целях и только с указанием авторства. Она включает следующие элементы:
* BY — обязательно указание авторства.
* NC — разрешено только некоммерческое использование.
CC-BY-NC и коммерческое использование допускаются по согласованию с авторами набора данных.
Справочные и дополнительные материалыИсходный набор данных
- https://github.com/snakers4/open_stt
Статьи на английском языке
- https://thegradient.pub/towards-an-imagenet-moment-for-speech-to-text/
- https://thegradient.pub/a-speech-to-text-practitioners-criticisms-of-industry-and-academia/
Статьи на китайском языке
- https://www.infoq.cn/article/4u58WcFCs0RdpoXev1E2
Статьи на русском языке
- https://habr.com/ru/post/494006/
- https://habr.com/ru/post/474462/
Access
Available in | When to use |
---|---|
Azure Notebooks | Quickly explore the dataset with Jupyter notebooks hosted on Azure or your local machine. |
Select your preferred service:
Azure Notebooks
!pip install numpy
!pip install tqdm
!pip install scipy
!pip install pandas
!pip install soundfile
!pip install librosa
!pip install azure-storage-blob
# manifest utils
import os
import numpy as np
import pandas as pd
from tqdm import tqdm
from urllib.request import urlopen
def reroot_manifest(manifest_df,
source_path,
target_path):
if source_path != '':
manifest_df.wav_path = manifest_df.wav_path.apply(lambda x: x.replace(source_path,
target_path))
manifest_df.text_path = manifest_df.text_path.apply(lambda x: x.replace(source_path,
target_path))
else:
manifest_df.wav_path = manifest_df.wav_path.apply(lambda x: os.path.join(target_path, x))
manifest_df.text_path = manifest_df.text_path.apply(lambda x: os.path.join(target_path, x))
return manifest_df
def save_manifest(manifest_df,
path,
domain=False):
if domain:
assert list(manifest_df.columns) == ['wav_path', 'text_path', 'duration', 'domain']
else:
assert list(manifest_df.columns) == ['wav_path', 'text_path', 'duration']
manifest_df.reset_index(drop=True).sort_values(by='duration',
ascending=True).to_csv(path,
sep=',',
header=False,
index=False)
return True
def read_manifest(manifest_path,
domain=False):
if domain:
return pd.read_csv(manifest_path,
names=['wav_path',
'text_path',
'duration',
'domain'])
else:
return pd.read_csv(manifest_path,
names=['wav_path',
'text_path',
'duration'])
def check_files(manifest_df,
domain=False):
orig_len = len(manifest_df)
if domain:
assert list(manifest_df.columns) == ['wav_path', 'text_path', 'duration']
else:
assert list(manifest_df.columns) == ['wav_path', 'text_path', 'duration', 'domain']
wav_paths = list(manifest_df.wav_path.values)
text_path = list(manifest_df.text_path.values)
omitted_wavs = []
omitted_txts = []
for wav_path, text_path in zip(wav_paths, text_path):
if not os.path.exists(wav_path):
print('Dropping {}'.format(wav_path))
omitted_wavs.append(wav_path)
if not os.path.exists(text_path):
print('Dropping {}'.format(text_path))
omitted_txts.append(text_path)
manifest_df = manifest_df[~manifest_df.wav_path.isin(omitted_wavs)]
manifest_df = manifest_df[~manifest_df.text_path.isin(omitted_txts)]
final_len = len(manifest_df)
if final_len != orig_len:
print('Removed {} lines'.format(orig_len-final_len))
return manifest_df
def plain_merge_manifests(manifest_paths,
MIN_DURATION=0.1,
MAX_DURATION=100):
manifest_df = pd.concat([read_manifest(_)
for _ in manifest_paths])
manifest_df = check_files(manifest_df)
manifest_df_fit = manifest_df[(manifest_df.duration>=MIN_DURATION) &
(manifest_df.duration<=MAX_DURATION)]
manifest_df_non_fit = manifest_df[(manifest_df.duration<MIN_DURATION) |
(manifest_df.duration>MAX_DURATION)]
print(f'Good hours: {manifest_df_fit.duration.sum() / 3600:.2f}')
print(f'Bad hours: {manifest_df_non_fit.duration.sum() / 3600:.2f}')
return manifest_df_fit
def save_txt_file(wav_path, text):
txt_path = wav_path.replace('.wav','.txt')
with open(txt_path, "w") as text_file:
print(text, file=text_file)
return txt_path
def read_txt_file(text_path):
#with open(text_path, 'r') as file:
response = urlopen(text_path)
file = response.readlines()
for i in range(len(file)):
file[i] = file[i].decode('utf8')
return file
def create_manifest_from_df(df, domain=False):
if domain:
columns = ['wav_path', 'text_path', 'duration', 'domain']
else:
columns = ['wav_path', 'text_path', 'duration']
manifest = df[columns]
return manifest
def create_txt_files(manifest_df):
assert 'text' in manifest_df.columns
assert 'wav_path' in manifest_df.columns
wav_paths, texts = list(manifest_df['wav_path'].values), list(manifest_df['text'].values)
# not using multiprocessing for simplicity
txt_paths = [save_txt_file(*_) for _ in tqdm(zip(wav_paths, texts), total=len(wav_paths))]
manifest_df['text_path'] = txt_paths
return manifest_df
def replace_encoded(text):
text = text.lower()
if '2' in text:
text = list(text)
_text = []
for i,char in enumerate(text):
if char=='2':
try:
_text.extend([_text[-1]])
except:
print(''.join(text))
else:
_text.extend([char])
text = ''.join(_text)
return text
# reading opus files
import os
import soundfile as sf
# Fx for soundfile read/write functions
def fx_seek(self, frames, whence=os.SEEK_SET):
self._check_if_closed()
position = sf._snd.sf_seek(self._file, frames, whence)
return position
def fx_get_format_from_filename(file, mode):
format = ''
file = getattr(file, 'name', file)
try:
format = os.path.splitext(file)[-1][1:]
format = format.decode('utf-8', 'replace')
except Exception:
pass
if format == 'opus':
return 'OGG'
if format.upper() not in sf._formats and 'r' not in mode:
raise TypeError("No format specified and unable to get format from "
"file extension: {0!r}".format(file))
return format
#sf._snd = sf._ffi.dlopen('/usr/local/lib/libsndfile/build/libsndfile.so.1.0.29')
sf._subtypes['OPUS'] = 0x0064
sf.SoundFile.seek = fx_seek
sf._get_format_from_filename = fx_get_format_from_filename
def read(file, **kwargs):
return sf.read(file, **kwargs)
def write(file, data, samplerate, **kwargs):
return sf.write(file, data, samplerate, **kwargs)
# display utils
import gc
from IPython.display import HTML, Audio, display_html
pd.set_option('display.max_colwidth', 3000)
#Prepend_path is set to read directly from Azure. To read from local replace below string with path to the downloaded dataset files
prepend_path = 'https://azureopendatastorage.blob.core.windows.net/openstt/ru_open_stt_opus_unpacked/'
def audio_player(audio_path):
return '<audio preload="none" controls="controls"><source src="{}" type="audio/wav"></audio>'.format(audio_path)
def display_manifest(manifest_df):
display_df = manifest_df
display_df['wav'] = [audio_player(prepend_path+path) for path in display_df.wav_path]
display_df['txt'] = [read_txt_file(prepend_path+path) for path in tqdm(display_df.text_path)]
audio_style = '<style>audio {height:44px;border:0;padding:0 20px 0px;margin:-10px -20px -20px;}</style>'
display_df = display_df[['wav','txt', 'duration']]
display(HTML(audio_style + display_df.to_html(escape=False)))
del display_df
gc.collect()
manifest_df = read_manifest(prepend_path +'/manifests/public_series_1.csv')
#manifest_df = reroot_manifest(manifest_df,
#source_path='',
#target_path='../../../../../nvme/stt/data/ru_open_stt/')
sample = manifest_df.sample(n=20)
display_manifest(sample)
!ls ru_open_stt_opus/manifests/*.csv
%matplotlib inline
import librosa
from scipy.io import wavfile
from librosa import display as ldisplay
from matplotlib import pyplot as plt
manifest_df = read_manifest(prepend_path +'manifests/asr_calls_2_val.csv')
#manifest_df = reroot_manifest(manifest_df,
#source_path='',
#target_path='../../../../../nvme/stt/data/ru_open_stt/')
sample = manifest_df.sample(n=5)
display_manifest(sample)
from io import BytesIO
wav_path = sample.iloc[0].wav_path
response = urlopen(prepend_path+wav_path)
data = response.read()
sr, wav = wavfile.read(BytesIO(data))
wav.astype('float32')
absmax = np.max(np.abs(wav))
wav = wav / absmax
# shortest way to plot a spectrogram
D = librosa.amplitude_to_db(np.abs(librosa.stft(wav)), ref=np.max)
plt.figure(figsize=(12, 6))
ldisplay.specshow(D, y_axis='log')
plt.colorbar(format='%+2.0f dB')
plt.title('Log-frequency power spectrogram')
# shortest way to plot an envelope
plt.figure(figsize=(12, 6))
ldisplay.waveplot(wav, sr=sr, max_points=50000.0, x_axis='time', offset=0.0, max_sr=1000, ax=None)
manifest_df = read_manifest(prepend_path +'manifests/asr_public_phone_calls_2.csv')
#manifest_df = reroot_manifest(manifest_df,
#source_path='',
#target_path='../../../../../nvme/stt/data/ru_open_stt/')
sample = manifest_df.sample(n=5)
display_manifest(sample)
opus_path = sample.iloc[0].wav_path
response = urlopen(prepend_path+opus_path)
data = response.read()
wav, sr = sf.read(BytesIO(data))
wav.astype('float32')
absmax = np.max(np.abs(wav))
wav = wav / absmax
# shortest way to plot a spectrogram
D = librosa.amplitude_to_db(np.abs(librosa.stft(wav)), ref=np.max)
plt.figure(figsize=(12, 6))
ldisplay.specshow(D, y_axis='log')
plt.colorbar(format='%+2.0f dB')
plt.title('Log-frequency power spectrogram')
# shortest way to plot an envelope
plt.figure(figsize=(12, 6))
ldisplay.waveplot(wav, sr=sr, max_points=50000.0, x_axis='time', offset=0.0, max_sr=1000, ax=None)