user_libri_text

  • Description:

UserLibri is a dataset containing paired audio-transcripts and additional text only data for each of 107 users. It is a reformatting of the LibriSpeech dataset found at http://www.openslr.org/12, reorganizing the data into users with an average of 52 LibriSpeech utterances and about 6,700 text example sentences per user. The UserLibriAudio class provides access to the audio-transcript pairs. See UserLibriText for the additional text data.

Split Examples
'10136' 38,496
'1041' 970
'10540' 3,283
'108' 5,864
'11' 1,348
'11667' 3,312
'1184' 22,062
'12176' 1,467
'12434' 2,796
'12544' 4,080
'13110' 2,634
'13158' 3,440
'13441' 4,145
'135' 37,263
'1353' 4,889
'1399' 18,914
'14420' 6,950
'14566' 3,810
'1477' 2,526
'14958' 1,495
'15263' 21,085
'15265' 7,647
'1549' 5,439
'1572' 2,882
'1597' 3,586
'1608' 3,605
'16127' 3,588
'16653' 7,600
'18096' 2,384
'1827' 4,806
'19019' 3,248
'19215' 13,542
'19717' 3,762
'1989' 1,105
'1998' 8,923
'20019' 966
'2002' 239
'20212' 3,363
'209' 2,090
'21297' 4,165
'22002' 4,044
'2300' 22,201
'24' 3,537
'24585' 1,789
'24811' 2,399
'2488' 8,239
'2529' 3,934
'26177' 3,598
'26379' 379
'2681' 8,872
'27067' 3,149
'27090' 3,217
'2770' 3,750
'2787' 4,603
'28700' 5,547
'28725' 3,899
'28952' 2,909
'2981' 54,305
'3076' 7,124
'30905' 2,140
'3178' 8,454
'33' 3,569
'33800' 5,145
'3436' 5,899
'3440' 5,087
'3441' 6,042
'36508' 521
'3748' 4,767
'38675' 2,696
'38804' 5,653
'39159' 2,729
'4028' 9,633
'40359' 7,821
'41326' 6,181
'4217' 6,003
'4276' 10,461
'434' 4,319
'4602' 4,421
'507' 9,093
'540' 5,452
'5516' 4,963
'5630' 1,130
'574' 452
'5921' 6,040
'6328' 5,926
'6812' 5,839
'732' 22,971
'76' 6,454
'7891' 1,476
'8166' 3,190
'820' 11,054
'833' 3,638
'9189' 8,387
'94' 1,722
'940' 6,172
'9464' 1,695
'955' 3,051
'969' 7,799
'9983' 8,898
  • Feature structure:
FeaturesDict({
    'book_id': Text(shape=(), dtype=tf.string),
    'text': Text(shape=(), dtype=tf.string),
})
  • Feature documentation:
Feature Class Shape Dtype Description
FeaturesDict
book_id Text tf.string The book that this text was pulled from
text Text tf.string A sentence of text extracted from a book
  • Citation:
@inproceedings{breiner2022userlibri,
  title={UserLibri: A Dataset for ASR Personalization Using Only Text},
  author={Breiner, Theresa and Ramaswamy, Swaroop and Variani, Ehsan and Garg, Shefali and Mathews, Rajiv and Sim, Khe Chai and Gupta, Kilol and Chen, Mingqing and McConnaughey, Lara},
  booktitle={Proc. Interspeech 2022},
  year={2022}
}