punctual-bert-bg

Visit the website - Zapetayko, to test out the model.

Usage

from transformers import pipeline


MODEL_ID = "auhide/punctual-bert-bg"

punctuate = pipeline("token-classification", model=MODEL_ID, tokenizer=MODEL_ID)
punctuate("Човекът искащ безгрижно писане ме помоли да създам този модел.")

[{'entity': 'B-CMA',
  'score': 0.95041466,
  'index': 1,
  'word': '▁Човекът',
  'start': 0,
  'end': 7},
 {'entity': 'I-CMA',
  'score': 0.95229745,
  'index': 2,
  'word': '▁иска',
  'start': 7,
  'end': 12},
 {'entity': 'B-CMA',
  'score': 0.95945585,
  'index': 5,
  'word': '▁писане',
  'start': 23,
  'end': 30},
 {'entity': 'I-CMA',
  'score': 0.90768945,
  'index': 6,
  'word': '▁ме',
  'start': 30,
  'end': 33}]

Basically, B-CMA tags the token that's before the comma, and I-CMA tags the token after the comma.

Therefore, if we place the commas based on these tags, the result is:

"Човекът, искащ безгрижно писане, ме помоли да създам този модел."