THCHS-30 - Aligned IPA transcriptions

This upload contains aligned IPA transcriptions for the THCHS-30 dataset from OpenSLR. Thereby, punctuation is added, silence marked and duration markers for each phoneme are assigned. Furthermore, the silence on the beginning and ending of each file is marked.

The words were transcribed using pypinyin (v0.47.1) via dict-from-pypinyin (v0.0.1) and mapped to IPA using the pinyin-ipa-map-TONE3-all.json mapping from here. The alignment was done using Montreal Forced Aligner (v2.0.5) and the acoustic model Mandarin MFA (v2.0.0a).

Phoneme duration markers:

˘ -> [0, 20) percentile (speaker-wise), e.g., a˥˩˘

(none) -> [20, 80) percentile (speaker-wise), e.g., a˥˩

ˑ -> [80, 90) percentile (speaker-wise), e.g., a˥˩ˑ

ː -> [90, inf) percentile (speaker-wise), e.g., a˥˩ː

Thereby each phoneme (including tones) was considered on its own, i.e., phonemes with different tones were not considered together for the percentile calculation.

Silence markers:

SILX -> silence at start/end of a recording (aligned on all tiers)

SIL0 -> no silence

SIL1 -> [0, 33.33333333) percentile of all silences (speaker-wise)

SIL2 -> [33.33333333, 66.66666666) percentile of all silences (speaker-wise)

SIL3 -> [66.66666666, inf) percentile of all silences (speaker-wise)

Files:

grids.zip

contains TextGrids for all audio files containing three tiers words, phonemes and transcription

words contains the aligned Chinese words

phonemes contains the IPA pronunciations including silence markers at start and end (SILX)

transcription contains unaligned phonemes including punctuation and word boundary labels (SIL0)

the folder structure is equal to one from the dataset

grids-sdp.zip

same as grids.zip except the folder structure is the one from speech-dataset-parser (v0.0.4)

preview-start/middle/end.png

preview of the first TextGrid from speaker A2 opened in Praat in different positions

words-vocabulary.txt

contains all Chinese words from tier words

phonemes-vocabulary.txt

contains all phonemes from tier phonemes

transcription-vocabulary.txt

contains all phonemes/punctuation from tier transcription

phonemes-durations.pdf

contains the plotted phoneme duration distribution of tier phonemes

phonemes-durations-simple.pdf

contains the plotted phoneme duration distribution of tier phonemes if all duration markers are ignored

phonemes-durations-simple-toneless.pdf

contains the plotted phoneme duration distribution of tier phonemes if all duration markers and tones are ignored

pronunciations-broad.dict

contains the broad pronunciations for each word including punctuation but not duration markers

e.g., 一下。 i˥ ɕ j a˥˩ 。

pronunciations-narrow.dict

contains the narrow pronunciations for each word including punctuation, duration markers and weights (= occurrence) over all speakers

e.g., 一下。 3 i˥ ɕ j a˥˩ː 。

pronunciations-narrow-speakers.zip

contains the narrow pronunciations separated for each speaker

script.sh

contains the script to reproduce all results

error in line 32: replace with speech-dataset-parser==0.0.4