MCPcopy Create free account
hub / github.com/SooLab/CGFormer / _run_split_on_punc

Method _run_split_on_punc

bert/tokenization_bert.py:414–434  ·  view source on GitHub ↗

Splits punctuation on a piece of text.

(self, text, never_split=None)

Source from the content-addressed store, hash-verified

412 return "".join(output)
413
414 def _run_split_on_punc(self, text, never_split=None):
415 """Splits punctuation on a piece of text."""
416 if never_split is not None and text in never_split:
417 return [text]
418 chars = list(text)
419 i = 0
420 start_new_word = True
421 output = []
422 while i < len(chars):
423 char = chars[i]
424 if _is_punctuation(char):
425 output.append([char])
426 start_new_word = True
427 else:
428 if start_new_word:
429 output.append([])
430 start_new_word = False
431 output[-1].append(char)
432 i += 1
433
434 return ["".join(x) for x in output]
435
436 def _tokenize_chinese_chars(self, text):
437 """Adds whitespace around any CJK character."""

Callers 1

tokenizeMethod · 0.95

Calls 1

_is_punctuationFunction · 0.85

Tested by

no test coverage detected