Pinyin - IPA Mapping

This upload contains IPA json-mappings for all pinyin romanized Chinese syllables retrieved from a big corpus.

The corpus that contained the original Chinese words was taken from uni-leipzig. We took the 1M Wikipedia Corpus from 2018. Each syllable was extracted and then converted to pinyin. The pinyin transcription was retrieved with pypinyin (v0.47.1) using dict-from-pypinyin (v0.0.1) which was then transcribed to IPA using pinyin-to-ipa (v0.0.1). Only the first possible transcription was included in the mappings.

Note: tone sandhi is not considered since the vocabulary consists only of stand-alone syllables.

Files:

• hanzi-vocabulary.txt
  • contains the hanzi vocabulary from which pinyin was transcribed (Chinese syllables), e.g., 㩳
• pinyin-ipa-map-NORMAL.json (418 mappings)
  • contains toneless pinyin mapped to IPA in pypinyin-style NORMAL, e.g., beng
• pinyin-ipa-map-TONE.json (1400 mappings)
  • contains pinyin mapped to IPA with pinyin tones in pypinyin-style TONE, e.g., bèng
• pinyin-ipa-map-TONE2.json (1400 mappings)
  • contains pinyin mapped to IPA with pinyin tones in pypinyin-style TONE2, e.g., be4ng
• pinyin-ipa-map-TONE3.json (1400 mappings)
  • contains pinyin mapped to IPA with pinyin tones in pypinyin-style TONE3, e.g., beng4
• pinyin-ipa-map-TONE3-all.json (2508 mappings)
  • contains all theoretical combinations of pinyin mapped to IPA with pinyin tones in pypinyin-style TONE3, e.g., beng4
• oov-vocabulary.txt
  • contains the vocabulary from which no pinyin could have been transcribed (because it was no Chinese symbol or doesn't have a pinyin representation), e.g., 방 or 㕔
• script.sh
  • contains the script to reproduce all results