Question

我有一个应该对文本文档列表进行预处理的代码。即：给定一个文本文档列表，它返回一个列表，其中每个文本文档都经过了预处理。但是由于某种原因，它无法删除标点符号。

import nltk
from nltk.tokenize import word_tokenize
from nltk.corpus import stopwords
nltk.download("stopwords")
nltk.download('punkt')
nltk.download('wordnet')


def preprocess(docs):
  """ 
  Given a list of documents, return each documents as a string of tokens, 
  stripping out punctuation 
  """
  clean_docs = [clean_text(i) for i in docs]
  tokenized_docs = [tokenize(i) for i in clean_docs]
  return tokenized_docs

def tokenize(text):
  """ 
  Tokenizes text -- returning the tokens as a string 
  """
  stop_words = stopwords.words("english")
  nltk_tokenizer = nltk.WordPunctTokenizer().tokenize
  tokens = nltk_tokenizer(text)  
  result = " ".join([i for i in tokens if not i in stop_words])
  return result


def clean_text(text): 
  """ 
  Cleans text by removing case
  and stripping out punctuation. 
  """
  new_text = make_lowercase(text)
  new_text = remove_punct(new_text)
  return new_text

def make_lowercase(text):
  new_text = text.lower()
  return new_text

def remove_punct(text):
  text = text.split()
  punct = string.punctuation
  new_text = " ".join(word for word in text if word not in string.punctuation)
  return new_text

# Get a list of titles  
s1 = "[UPDATE] I am tired"
s2 = "I am cold."

clean_docs = preprocess([s1, s2])
print(clean_docs)

打印输出：

['[ update ] tired', 'cold .']

换句话说，它不会删除标点符号，因为“ [”，“]”和“。”全部出现在最终产品中。

Answer 1

您要搜索标点中的单词。显然[UPDATE]不是标点符号。

尝试在文本中搜索标点/替换标点：

import string


def remove_punctuation(text: str) -> str:
    for p in string.punctuation:
        text = text.replace(p, '')
    return text


if __name__ == '__main__':
    text = '[UPDATE] I am tired'
    print(remove_punctuation(text))

# output:
# UPDATE I am tired

预处理脚本无法删除标点符号

1 个答案: