nlp - 如何在 Stanford CoreNLP 中获取短语标签？

Question

如果我想获取每个单词对应的短语标签，我该如何获取？

例如:

在这句话中，

My dog also likes eating sausage.

我可以在 Stanford NLP 中得到一个解析树，例如

(ROOT (S (NP (PRP$ My) (NN dog)) (ADVP (RB also)) (VP (VBZ likes) (NP (JJ eating) (NN sausage))) (. .)))

在上面的情况下，我想得到对应每个单词的短语标签

(My - NP), (dog - NP), (also - ADVP), (likes - VP), ...

有什么简单的词组标签提取方法吗？

请帮帮我。

Answer 1

//I guess this is how you get your parse tree.
Tree tree = sentAnno.get(TreeAnnotation.class);

//The children of a Tree annotation is an array of trees.
Tree[] children = parent.children() 

//Check the label of any sub tree to see whether it is what you want (a phrase)
for (Tree child: children){
   if (child.value().equals("NP")){// set your rule of defining Phrase here
          List<Tree> leaves = child.getLeaves(); //leaves correspond to the tokens
          for (Tree leaf : leaves){ 
            List<Word> words = leaf.yieldWords();
            for (Word word: words)
                System.out.print(String.format("(%s - NP),",word.word()));
          }
   }
}

该代码未经过全面测试，但我认为它大致可以满足您的需求。更重要的是，我没有写任何关于递归访问子树的内容，但我相信你应该能够做到这一点。