from vnbarefixing.bigram_fixing import BigramFixing
bigram_fixing = BigramFixing.load()
sen = "Sentence tobe fixed"
fixed_bare_sen,ref_accent_sen = bigram_fixing.fix(sen)

Dir: models/data

• full_vocabulary.txt: Vocaburaly file

Dir: models/data/inp

• new_fixing: contains new fixs, this file is reloaded each time the fixer is initialized without retraining. The syntax of each line:

  (wrong word)\t(true word)

• common_fixing: contains common fixing with the syntax of each line:

   (wrong word)\t(true word)

• hard_regexes: contains hard regexes for replacing:

    (pattern)\t(sub)

• missing_vocab: contains missing words of the vocabulary
• special_tokens: contains special single token
• speical_words: contains special words

from vnbarefixing.bigram_fixing import BigramFixing
BigramFixing.train(data="",path="")
#data is a list of unicode sentences
#path specifies the file saving sentences
#only select one of them