ACV_chn_beta_0.0.1

SOFA model trained on ACV style mandarin Chinese phonemes. Annotated by hand (and manually adjusted output from WFL), including data of ACV-001 and a small portion of OpenCPop (not fully relabeled, hence only a portion). Currently this is just a beta version due to the lack of data diversity, but will hopefully be updated once more data is obtained.

A dictionary is included for inference (dict_cn.txt)

Datasets used:

Full label:
- ACV-001
- Partial OpenCPop (only around 3 songs)

Weak label:   # The same dataset as above, except the `ph_dur` is removed from the transcriptions.csv
- ACV-001
- Partial OpenCPop

No label:    # Again using the same dataset, this time without transcriptions.csv
- ACV-001
- Partial OpenCPop