如何使用Python从句子中提取名词性复合词？

下面提供几种实用方法，帮你用NLTK提取“basketball shoe”这类名词组合：

• 方法1：自定义正则分块器
  通过定义规则匹配连续的名词词性标签，将它们合并为名词短语：

  import nltk
  from nltk.chunk import RegexpParser
  
  sentence = "I bought a basketball shoe and a coffee table yesterday"
  words = nltk.word_tokenize(sentence)
  tags = nltk.pos_tag(words)
  
  # 规则：匹配任意数量连续的名词标签（NN、NNP、NNS等）
  chunk_rule = r"""
  NP: {<NN.*>+}
  """
  chunk_parser = RegexpParser(chunk_rule)
  parsed_result = chunk_parser.parse(tags)
  
  # 提取所有名词短语
  noun_phrases = []
  for subtree in parsed_result.subtrees(filter=lambda t: t.label() == 'NP'):
      phrase = ' '.join(word for word, tag in subtree.leaves())
      noun_phrases.append(phrase)
  
  print(noun_phrases)  # 输出: ['basketball shoe', 'coffee table']
  

• 方法2：使用预训练分块模型
  NLTK自带基于树库训练的分块器，能处理更复杂的名词短语结构：

  import nltk
  from nltk.corpus import treebank_chunk
  
  # 下载并加载预训练分块器
  nltk.download('treebank_chunk')
  chunker = treebank_chunk.chunkers()[0]
  
  sentence = "My favorite sports gear is basketball shoe and tennis racket"
  words = nltk.word_tokenize(sentence)
  tags = nltk.pos_tag(words)
  
  parsed_result = chunker.parse(tags)
  noun_phrases = []
  for subtree in parsed_result.subtrees(filter=lambda t: t.label() == 'NP'):
      phrase = ' '.join(word for word, tag in subtree.leaves())
      noun_phrases.append(phrase)
  
  print(noun_phrases)
  

• 方法3：滑动窗口合并连续名词
  逻辑简单直接，遍历词性序列，把连续的名词标签对应的词合并：

  import nltk
  
  sentence = "She placed the wooden dining table near the window"
  words = nltk.word_tokenize(sentence)
  tags = nltk.pos_tag(words)
  
  noun_phrases = []
  current_phrase = []
  for word, tag in tags:
      if tag.startswith('NN'):
          current_phrase.append(word)
      else:
          if current_phrase:
              noun_phrases.append(' '.join(current_phrase))
              current_phrase = []
  # 处理最后一组未闭合的名词
  if current_phrase:
      noun_phrases.append(' '.join(current_phrase))
  
  print(noun_phrases)  # 输出: ['wooden dining table', 'window']
  

内容的提问来源于stack exchange，提问作者Alan K