Could not find where 'punct_default' is imported

[kathy-lee]

Hi dear community,

can someone kindly tell me where the variable punct_default from fun_text_processing/text_normalization /data_loader_utils.py file is imported? I could not find in the whole repo. And what does it mean? https://github.com/modelscope/FunASR/blob/5ee2f382b32c5c0beaa1c133b466a100c6fb7ebc/fun_text_processing/text_normalization/data_loader_utils.py#L302

[LauraGPT]

You were right: punct_default was never imported or defined. It was a latent bug in the add_unicode_punct=True path, not a symbol supplied elsewhere.

The same block had two more defects:

1. sys was also not imported, so execution failed there first.
2. list.extend() returns None, but its return value was assigned back to punct_marks.

All three problems are fixed in merged PR #3238, commit 1d8080af. The implementation now inspects only unique characters present in the input, selects Unicode category P characters that are not already covered by string.punctuation, and extends the punctuation list in place. This also avoids scanning the entire Unicode range for every call.

The fix includes regression coverage for both the unchanged ASCII behavior and Chinese Unicode punctuation. Please update to current main or a future release containing 1d8080a.