Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

JSUT corpus: free large-scale Japanese speech corpus for end-to-end speech synthesis

Ryosuke Sonobe, Shinnosuke Takamichi, Hiroshi Saruwatari

TL;DR

The paper presents the JSUT corpus, a free, large-scale 10-hour Japanese speech dataset designed to support end-to-end text-to-speech research. It achieves broad pronunciation coverage by assembling nine specialized sub-corpora that collectively address daily-use kanji readings, numerals with counters, loanwords, paraphrases, voice-actress para-speech, onomatopoeia, and domain-specific content. The authors describe recording details (48 kHz, 16-bit WAV, UTF-8 transcripts) and provide initial analyses of utterance lengths and pitch statistics, validating diversity across days. The JSUT corpus enables research across academia and industry by providing a practical, multi-domain resource for building Japanese end-to-end TTS systems.

Abstract

Thanks to improvements in machine learning techniques including deep learning, a free large-scale speech corpus that can be shared between academic institutions and commercial companies has an important role. However, such a corpus for Japanese speech synthesis does not exist. In this paper, we designed a novel Japanese speech corpus, named the "JSUT corpus," that is aimed at achieving end-to-end speech synthesis. The corpus consists of 10 hours of reading-style speech data and its transcription and covers all of the main pronunciations of daily-use Japanese characters. In this paper, we describe how we designed and analyzed the corpus. The corpus is freely available online.

JSUT corpus: free large-scale Japanese speech corpus for end-to-end speech synthesis

TL;DR

The paper presents the JSUT corpus, a free, large-scale 10-hour Japanese speech dataset designed to support end-to-end text-to-speech research. It achieves broad pronunciation coverage by assembling nine specialized sub-corpora that collectively address daily-use kanji readings, numerals with counters, loanwords, paraphrases, voice-actress para-speech, onomatopoeia, and domain-specific content. The authors describe recording details (48 kHz, 16-bit WAV, UTF-8 transcripts) and provide initial analyses of utterance lengths and pitch statistics, validating diversity across days. The JSUT corpus enables research across academia and industry by providing a practical, multi-domain resource for building Japanese end-to-end TTS systems.

Abstract

Thanks to improvements in machine learning techniques including deep learning, a free large-scale speech corpus that can be shared between academic institutions and commercial companies has an important role. However, such a corpus for Japanese speech synthesis does not exist. In this paper, we designed a novel Japanese speech corpus, named the "JSUT corpus," that is aimed at achieving end-to-end speech synthesis. The corpus consists of 10 hours of reading-style speech data and its transcription and covers all of the main pronunciations of daily-use Japanese characters. In this paper, we describe how we designed and analyzed the corpus. The corpus is freely available online.

Paper Structure

This paper contains 16 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Corpus design
  3. Structures
  4. Components
  5. basic5000
  6. countersuffix26
  7. loanword128
  8. utparaphrase512
  9. voiceactress100
  10. Onomatopee300
  11. repeat500
  12. travel1000 and precedent138
  13. Results of data collection
  14. Corpus specs
  15. Analysis
  16. ...and 1 more sections

Figures (3)

  • Figure 1: Histogram of number of moras (sub-syllables) in one utterance. Minimum, mean, and maximum values are 7, 37.14, and 133, respectively.
  • Figure 2: Histogram of number of words in one utterance. Minimum, mean, and maximum values are 2, 18.03, and 70, respectively.
  • Figure 3: Mean of log-scaled F0 for each recording day. Ordinal number of x-axis means how much time passed from "1st" recording day. For example, "5th" means 4 days after 1st recording day.