NLTK：如何遍歷名詞短語以返回字符串列表？

Question

在 NLTK 中，如何遍歷已解析的句子以返回名詞短語字符串列表？

我有兩個目標：
(1) 創建名詞短語列表而不是使用“traverse()”方法打印它們。 我目前使用StringIO來記錄現有的traverse()方法的output。 這不是一個可接受的解決方案。
(2) 反解析名詞短語字符串，這樣：'(NP Michael/NNP Jackson/NNP)' 變成'Michael Jackson'。 NLTK 中有反解析的方法嗎？

NLTK 文檔建議使用 traverse() 來查看名詞短語，但是如何在此遞歸方法中捕獲“t”以便生成字符串名詞短語列表？

from nltk.tag import pos_tag

def traverse(t):
  try:
      t.label()
  except AttributeError:
      return
  else:
      if t.label() == 'NP': print(t)  # or do something else
      else:
          for child in t: 
              traverse(child)

def nounPhrase(tagged_sent):
    # Tag sentence for part of speech
    tagged_sent = pos_tag(sentence.split())  # List of tuples with [(Word, PartOfSpeech)]
    # Define several tag patterns
    grammar = r"""
      NP: {<DT|PP\$>?<JJ>*<NN>}   # chunk determiner/possessive, adjectives and noun
      {<NNP>+}                # chunk sequences of proper nouns
      {<NN>+}                 # chunk consecutive nouns
      """
    cp = nltk.RegexpParser(grammar)  # Define Parser
    SentenceTree = cp.parse(tagged_sent)
    NounPhrases = traverse(SentenceTree)   # collect Noun Phrase
    return(NounPhrases)

sentence = "Michael Jackson likes to eat at McDonalds"
tagged_sent = pos_tag(sentence.split())  
NP = nounPhrase(tagged_sent)  
print(NP)  

這目前打印：
（NP 邁克爾/NNP 傑克遜/NNP）
（NP 麥當勞/NNP）
並將“無”存儲到 NP

Answer 1

def extract_np(psent):
  for subtree in psent.subtrees():
    if subtree.label() == 'NP':
      yield ' '.join(word for word, tag in subtree.leaves())


cp = nltk.RegexpParser(grammar)
parsed_sent = cp.parse(tagged_sent)
for npstr in extract_np(parsed_sent):
    print (npstr)

Answer 2

提取名詞短語的另一種可能性是使用Constituent-Treelib 庫，它可以通過以下方式安裝： pip install constituent-treelib 。

使用這個庫，我們需要執行以下步驟來提取（名詞）短語：

from constituent_treelib import ConstituentTree, BracketedTree

# First, we define the parsed sentence from where we want to extract phrases
parsed_sentence = "(S (NP (NNP Michael) (NNP Jackson)) (VP (VBZ likes) (S (VP (TO to) (VP (VB eat) (PP (IN at) (NP (NNPS McDonalds))))))))"

# ...and wrap the parsed sentence into a BracketedTree object
parsed_sentence = BracketedTree(parsed_sentence)

# Next, we define the language that should be considered with respect to the underlying models 
language = ConstituentTree.Language.English

# You can also specify the desired model for the language ("Small" is selected by default)
spacy_model_size = ConstituentTree.SpacyModelSize.Large

# Now, we create the neccesary NLP pipeline, which is required to create a ConstituentTree object
nlp = ConstituentTree.create_pipeline(language, spacy_model_size) 

# If you wish, you can instruct the library to download and install the models automatically
# nlp = ConstituentTree.create_pipeline(language, spacy_model_size, download_models=True) 

# Now we can instantiate a ConstituentTree object and pass it the parsed sentence as well as the NLP pipeline
tree = ConstituentTree(parsed_sentence, nlp)

# Finally, we can extract all phrases from the tree  
all_phrases = tree.extract_all_phrases(avoid_nested_phrases=True) 

>>> {'S': ['Michael Jackson likes to eat at McDonalds'],
>>> 'NP': ['Michael Jackson'],
>>> 'VP': ['likes to eat at McDonalds'],
>>> 'PP': ['at McDonalds']}

# ...or restrict them only to noun phrases
noun_phrases = all_phrases['NP']

>>> ['Michael Jackson']

如果您還想可視化樹，可以按如下方式進行：

tree.export_tree('my_tree.pdf')

結果：