[BUG]

[zihangli-hndsm]

A bug appeared when running the aligner on Mandarin Chinese corpus. The g2p didn't function properly.

1. Corpus structure
   The corpus is a Chinese Mandarin corpus, with 42 speakers, and I was using TextGrid files for input.

2. Dictionary
   I am using the default dictionary "mandarin_mfa".

3. Acoustic model
   I am using the default model "mandarin_mfa".

The log file is in the attachments.
try3.log

System: Ubuntu 22.04.2 LTS (GNU/Linux 5.15.153.1-microsoft-standard-WSL2 x86_64)

Details:
Some characters (about 30%) that are in the dictionary wasn't tagged as their corresponding IPA symbols, but "spn".

If you need more information, please let me know! Thank you very much!