4.缩写词

contraction.py 折叠源码

# -*- coding: utf-8 -*- """ Created on Mon Aug 01 01:11:02 2016 @author: DIP """   CONTRACTION_MAP = { "ain't": "is not", "aren't": "are not", "can't": "cannot", "can't've": "cannot have", "'cause": "because", "could've": "could have", "couldn't": "could not", "couldn't've": "could not have", "didn't": "did not", "doesn't": "does not", "don't": "do not", "hadn't": "had not", "hadn't've": "had not have", "hasn't": "has not", "haven't": "have not", "he'd": "he would", "he'd've": "he would have", "he'll": "he will", "he'll've": "he he will have", "he's": "he is", "how'd": "how did", "how'd'y": "how do you", "how'll": "how will", "how's": "how is", "I'd": "I would", "I'd've": "I would have", "I'll": "I will", "I'll've": "I will have", "I'm": "I am", "I've": "I have", "i'd": "i would", "i'd've": "i would have", "i'll": "i will", "i'll've": "i will have", "i'm": "i am", "i've": "i have", "isn't": "is not", "it'd": "it would", "it'd've": "it would have", "it'll": "it will", "it'll've": "it will have", "it's": "it is", "let's": "let us", "ma'am": "madam", "mayn't": "may not", "might've": "might have", "mightn't": "might not", "mightn't've": "might not have", "must've": "must have", "mustn't": "must not", "mustn't've": "must not have", "needn't": "need not", "needn't've": "need not have", "o'clock": "of the clock", "oughtn't": "ought not", "oughtn't've": "ought not have", "shan't": "shall not", "sha'n't": "shall not", "shan't've": "shall not have", "she'd": "she would", "she'd've": "she would have", "she'll": "she will", "she'll've": "she will have", "she's": "she is", "should've": "should have", "shouldn't": "should not", "shouldn't've": "should not have", "so've": "so have", "so's": "so as", "that'd": "that would", "that'd've": "that would have", "that's": "that is", "there'd": "there would", "there'd've": "there would have", "there's": "there is", "they'd": "they would", "they'd've": "they would have", "they'll": "they will", "they'll've": "they will have", "they're": "they are", "they've": "they have", "to've": "to have", "wasn't": "was not", "we'd": "we would", "we'd've": "we would have", "we'll": "we will", "we'll've": "we will have", "we're": "we are", "we've": "we have", "weren't": "were not", "what'll": "what will", "what'll've": "what will have", "what're": "what are", "what's": "what is", "what've": "what have", "when's": "when is", "when've": "when have", "where'd": "where did", "where's": "where is", "where've": "where have", "who'll": "who will", "who'll've": "who will have", "who's": "who is", "who've": "who have", "why's": "why is", "why've": "why have", "will've": "will have", "won't": "will not", "won't've": "will not have", "would've": "would have", "wouldn't": "would not", "wouldn't've": "would not have", "y'all": "you all", "y'all'd": "you all would", "y'all'd've": "you all would have", "y'all're": "you all are", "y'all've": "you all have", "you'd": "you would", "you'd've": "you would have", "you'll": "you will", "you'll've": "you will have", "you're": "you are", "you've": "you have" }

以下 代码段载入英文停用词、 CONTRACTION_MAP 缩写词汇表、 WordNetLemmatizer 的一个实例

from contractions import CONTRACTION_MAP import re import nltk import string from nltk.stem import WordNetLemmatizer   stopword_list = nltk.corpus.stopwords.words('english') wnl = WordNetLemmatizer()

以下是词语切分函数，并去除分割后的多余空格。

def tokenize_text(text):     tokens = nltk.word_tokenize(text)     tokens = [token.strip() for token in tokens]     return tokens

以下是扩展缩写词的函数，返回包含扩展缩写词后的文本：

def expand_contractions(text, contraction_mapping):     contractions_pattern = re.compile('({})'.format('|'.join(contraction_mapping.keys())),                                       flags=re.IGNORECASE|re.DOTALL)     def expand_match(contraction):         match = contraction.group(0)         first_char = match[0]         expanded_contraction = contraction_mapping.get(match)\                                 if contraction_mapping.get(match)\                                 else contraction_mapping.get(match.lower())                               expanded_contraction = first_char+expanded_contraction[1:]         return expanded_contraction     expanded_text = contractions_pattern.sub(expand_match, text)     return expanded_text