为什么在使用“nltk.ne_chunk”分块后我没有得到“PERSON”和“GPE”作为标签？

Question

我正在像这样使用nltk.ne_chunk()：

sent="Azhar is asking what is weather in Chicago today? "
chunks = nltk.ne_chunk(nltk.pos_tag(nltk.word_tokenize(sent)), binary=True)
print(list(chunks))

然后像这样得到 oitput：

[Tree('NE', [('Azhar', 'NNP')]), ('is', 'VBZ'), ('asking', 'VBG'), ('what', 'WP'), ('is', 
'VBZ'), ('weather', 'NN'), ('in', 'IN'), Tree('NE', [('Chicago', 'NNP')]), ('today', 'NN'), 
('?', '.')]

但我期待这样的输出：

[Tree('PERSON', [('Azhar', 'NNP')]), ('is', 'VBZ'), ('asking', 'VBG'), ('what', 'WP'), ('is', 
'VBZ'), ('weather', 'NN'), ('in', 'IN'), Tree('GPE', [('Chicago', 'NNP')]), ('today', 'NN'), 
('?', '.')]

谁能告诉我我在这里做错了什么？

Answer 1

安装Spacy库并下载相关模型（en_core_web_sm）后，解释here，您可以简单地提取Named-Entities！

import spacy
NER = spacy.load("en_core_web_sm")
sent="Azhar is asking what is weather in Chicago today? "
text1= NER(sent)
for word in text1.ents:
    print(word.text,word.label_)

输出：

Azhar PERSON
Chicago GPE
today DATE

更新

nltk.ne_chunk 返回一个嵌套的nltk.tree.Tree 对象，因此您必须遍历 Tree 对象才能到达 NE。来自nltk.chunk 的tree2conlltags 会这样做！

from nltk import word_tokenize, pos_tag, ne_chunk
from nltk.chunk import tree2conlltags

sentence = "Azhar is asking what is weather in Chicago today?"
print(tree2conlltags(ne_chunk(pos_tag(word_tokenize(sentence)))))

以 IOB 格式输出：

[('Azhar', 'NNP', 'B-GPE'), ('is', 'VBZ', 'O'), ('asking', 'VBG', 'O'), ('what', 'WP', 'O'), ('is', 'VBZ', 'O'), ('weather', 'NN', 'O'), ('in', 'IN', 'O'), ('Chicago', 'NNP', 'B-GPE'), ('today', 'NN', 'O'), ('?', '.', 'O')]

更多关于这个here！