Mandarin syllable-level (pinyin) alignment

[kc-li]

I'm been using the latest Mandarin acoustic model Mandarin MFA acoustic model v3.0.0 together with the Mandarin (China) MFA dictionary v3.0.0.

I really appreciate the amazing accuracy of the aligned results! That said, I'm hoping to get character-by-character (i.e. syllable-by-syllable) boundaries from the output. At the moment, achieving this seems to require inferring syllable boundaries from the ipa strings, which could be done through some algorithms. But before going down that route, I wanted to check if there are recommended approaches within MFA that can help me achieve that goal more directly.

1. I noticed that there are existing models and dictionaries using pinyin: Mandarin (China Pinyin) MFA G2P model v3.0.0 and Mandarin PINYIN dictionary v2.0.0. The pinyin alignment would make it easier to achieve my goal as the segmentation only distinguishes onset and rhyme, not individual segment. Would it be possible to make use of these pinyin-based resources, while still using the latest acoustic model?

2. As an alternative, could I modify the dictionary directly to force syllable-level alignment?
   For example, if I take the Mandarin (China) MFA dictionary v3.0.0 and strip it down to only mono-syllabic entries, would that encourage the aligner to segment and align the input syllable by syllable?

Any suggestion would be helpful!

[kalimandfas]

mybe you can try use https://github.com/GalaxieT/MFA-mandarin-pinyin-dict-for-pretrained-mfa-model-v2.0,it can use pinyin and mfav2 model,but i am not sure if it is not useful