lj_speech / README.md
albertvillanova's picture
Revert Convert dataset to Parquet (#6)
aa96d9a verified
metadata
annotations_creators:
  - expert-generated
language_creators:
  - found
language:
  - en
license:
  - unlicense
multilinguality:
  - monolingual
paperswithcode_id: ljspeech
pretty_name: LJ Speech
size_categories:
  - 10K<n<100K
source_datasets:
  - original
task_categories:
  - automatic-speech-recognition
  - text-to-speech
  - text-to-audio
task_ids: []
train-eval-index:
  - config: main
    task: automatic-speech-recognition
    task_id: speech_recognition
    splits:
      train_split: train
    col_mapping:
      file: path
      text: text
    metrics:
      - type: wer
        name: WER
      - type: cer
        name: CER
dataset_info:
  features:
    - name: id
      dtype: string
    - name: audio
      dtype:
        audio:
          sampling_rate: 22050
    - name: file
      dtype: string
    - name: text
      dtype: string
    - name: normalized_text
      dtype: string
  config_name: main
  splits:
    - name: train
      num_bytes: 4667022
      num_examples: 13100
  download_size: 2748572632
  dataset_size: 4667022

Dataset Card for lj_speech

Table of Contents

Dataset Description

Dataset Summary

This is a public domain speech dataset consisting of 13,100 short audio clips of a single speaker reading passages from 7 non-fiction books in English. A transcription is provided for each clip. Clips vary in length from 1 to 10 seconds and have a total length of approximately 24 hours.

The texts were published between 1884 and 1964, and are in the public domain. The audio was recorded in 2016-17 by the LibriVox project and is also in the public domain.

Supported Tasks and Leaderboards

The dataset can be used to train a model for Automatic Speech Recognition (ASR) or Text-to-Speech (TTS).

  • automatic-speech-recognition: An ASR model is presented with an audio file and asked to transcribe the audio file to written text. The most common ASR evaluation metric is the word error rate (WER).
  • text-to-speech, text-to-audio: A TTS model is given a written text in natural language and asked to generate a speech audio file. A reasonable evaluation metric is the mean opinion score (MOS) of audio quality. The dataset has an active leaderboard which can be found at https://paperswithcode.com/sota/text-to-speech-synthesis-on-ljspeech

Languages

The transcriptions and audio are in English.

Dataset Structure

Data Instances

A data point comprises the path to the audio file, called file and its transcription, called text. A normalized version of the text is also provided.

{
    'id': 'LJ002-0026',
    'file': '/datasets/downloads/extracted/05bfe561f096e4c52667e3639af495226afe4e5d08763f2d76d069e7a453c543/LJSpeech-1.1/wavs/LJ002-0026.wav',
    'audio': {'path': '/datasets/downloads/extracted/05bfe561f096e4c52667e3639af495226afe4e5d08763f2d76d069e7a453c543/LJSpeech-1.1/wavs/LJ002-0026.wav',
      'array': array([-0.00048828, -0.00018311, -0.00137329, ...,  0.00079346,
              0.00091553,  0.00085449], dtype=float32),
      'sampling_rate': 22050},
    'text': 'in the three years between 1813 and 1816,'
    'normalized_text': 'in the three years between eighteen thirteen and eighteen sixteen,',
}

Each audio file is a single-channel 16-bit PCM WAV with a sample rate of 22050 Hz.

Data Fields

  • id: unique id of the data sample.

  • file: a path to the downloaded audio file in .wav format.

  • audio: A dictionary containing the path to the downloaded audio file, 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].

  • text: the transcription of the audio file.

  • normalized_text: the transcription with numbers, ordinals, and monetary units expanded into full words.

Data Splits

The dataset is not pre-split. Some statistics:

  • Total Clips: 13,100
  • Total Words: 225,715
  • Total Characters: 1,308,678
  • Total Duration: 23:55:17
  • Mean Clip Duration: 6.57 sec
  • Min Clip Duration: 1.11 sec
  • Max Clip Duration: 10.10 sec
  • Mean Words per Clip: 17.23
  • Distinct Words: 13,821

Dataset Creation

Curation Rationale

[Needs More Information]

Source Data

Initial Data Collection and Normalization

This dataset consists of excerpts from the following works:

  • Morris, William, et al. Arts and Crafts Essays. 1893.
  • Griffiths, Arthur. The Chronicles of Newgate, Vol. 2. 1884.
  • Roosevelt, Franklin D. The Fireside Chats of Franklin Delano Roosevelt. 1933-42.
  • Harland, Marion. Marion Harland's Cookery for Beginners. 1893.
  • Rolt-Wheeler, Francis. The Science - History of the Universe, Vol. 5: Biology. 1910.
  • Banks, Edgar J. The Seven Wonders of the Ancient World. 1916.
  • President's Commission on the Assassination of President Kennedy. Report of the President's Commission on the Assassination of President Kennedy. 1964.

Some details about normalization:

  • The normalized transcription has the numbers, ordinals, and monetary units expanded into full words (UTF-8)
  • 19 of the transcriptions contain non-ASCII characters (for example, LJ016-0257 contains "raison d'être").
  • The following abbreviations appear in the text. They may be expanded as follows:
Abbreviation Expansion
Mr. Mister
Mrs. Misess (*)
Dr. Doctor
No. Number
St. Saint
Co. Company
Jr. Junior
Maj. Major
Gen. General
Drs. Doctors
Rev. Reverend
Lt. Lieutenant
Hon. Honorable
Sgt. Sergeant
Capt. Captain
Esq. Esquire
Ltd. Limited
Col. Colonel
Ft. Fort
(*) there's no standard expansion for "Mrs."

Who are the source language producers?

[Needs More Information]

Annotations

Annotation process

  • The audio clips range in length from approximately 1 second to 10 seconds. They were segmented automatically based on silences in the recording. Clip boundaries generally align with sentence or clause boundaries, but not always.
  • The text was matched to the audio manually, and a QA pass was done to ensure that the text accurately matched the words spoken in the audio.

Who are the annotators?

Recordings by Linda Johnson from LibriVox. Alignment and annotation by Keith Ito.

Personal and Sensitive Information

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.

Considerations for Using the Data

Social Impact of Dataset

[Needs More Information]

Discussion of Biases

[Needs More Information]

Other Known Limitations

  • The original LibriVox recordings were distributed as 128 kbps MP3 files. As a result, they may contain artifacts introduced by the MP3 encoding.

Additional Information

Dataset Curators

The dataset was initially created by Keith Ito and Linda Johnson.

Licensing Information

Public Domain (LibriVox)

Citation Information

@misc{ljspeech17,
  author       = {Keith Ito and Linda Johnson},
  title        = {The LJ Speech Dataset},
  howpublished = {\url{https://keithito.com/LJ-Speech-Dataset/}},
  year         = 2017
}

Contributions

Thanks to @anton-l for adding this dataset.