CoolFace
Datasetpublic

changelinglab/thchs30-segment

THCHS-30 Segment Mandarin Chinese read-speech corpus with phone-level time alignments. Suitable for training and evaluating phone recognition and phonetic segmentation models. Sources Audio: THCHS-30 (OpenSLR 18) by Dong Wang, Xuewei Zhang, Zhiyong Zhang (Tsinghua University, 2015). Phone alignments: anyspeech/THCHS-30-alignments. Splits Split Utterances train 10,000 val 893 test 2,495 Splits follow the original OpenSLR 18… See the full description on the dataset page: https://huggingface.co/datasets/changelinglab/thchs30-segment.

sourceHugging Faceapache-2.0updated 6mo agoView on Hugging Face
0likes150downloads
Dataset Card

THCHS-30 Segment

Mandarin Chinese read-speech corpus with phone-level time alignments. Suitable for training and evaluating phone recognition and phonetic segmentation models.

Sources

Splits

SplitUtterances
train10,000
val893
test2,495

Splits follow the original OpenSLR 18 directory partition (data_thchs30/{train,dev,test}); dev is renamed to val.

Schema

ColumnTypeDescription
utt_idstringUtterance id, e.g. A11_0
audioAudio(16 kHz)Embedded waveform bytes (decoded on access)
textstringHanzi sentence transcript
phonessequence[string]IPA phone tokens with tone diacritics
phone_startssequence[float64]Phone start times in seconds
phone_endssequence[float64]Phone end times in seconds
languagestringcmn (ISO 639-3)
speaker_idstringSpeaker code (utt_id prefix, e.g. A11)
durationfloat64Utterance duration in seconds
splitstringtrain / val / test

Phone inventory

Phones are IPA with Mandarin tone diacritics, e.g. lː, y˥˩, ʂ˘, ɻ̩˥˩, a˧˥˘. Silence and pauses are marked with [SIL] intervals, which are kept in the alignment so boundary models can learn from them.

License

Released under the Apache 2.0 license, matching the original THCHS-30 release.

Citation

bibtex
@misc{THCHS30_2015,
  title={THCHS-30 : A Free Chinese Speech Corpus},
  author={Dong Wang, Xuewei Zhang, Zhiyong Zhang},
  year={2015},
  url={http://arxiv.org/abs/1512.01882}
}