我生成这个NLTK语法的方式有问题吗？

Question

我用NLTK编写了这个简单的程序，它只是打印出语法树。但是，即使正在创建RecursiveDescentParser，它也不会输出任何内容。我有什么问题？我对语法的定义有误吗？我试图迭代解析器的方式有问题吗？提前谢谢你。

import nltk

'''The price of peace is rising.'''

grammar = nltk.CFG.fromstring("""
  S -> NP VP
  VP -> V NP | V NP PP
  PP -> P NP
  V -> "is" | "rising"
  NP -> Det N | Det N PP
  Det -> "the" | "of"
  N -> "price" | "peace"
  P -> "in" | "on" | "by" | "with"
  """)

sentence = "the price of peace is rising"
wordArray = sentence.split()

print(wordArray)

parser = nltk.RecursiveDescentParser(grammar)

for tree in parser.parse(wordArray):
    print(tree)

Answer 1

首先，一定要从一口一口开始写语法。

让我们从一个简单的句子Peace is rising开始。

我们需要结构S -> NP VP，其中：

• VP是一个不及物动词短语，在这种情况下，is rising与辅助is连用，rise与-ing进行词尾变化连用。
• NP只是一个单独的名词。

代码

import nltk

your_grammar = nltk.CFG.fromstring("""
S -> NP VP
VP -> AUX V 
V -> "rising"
AUX -> "is"
NP -> N
N -> "peace"
""")

parser = nltk.RecursiveDescentParser(your_grammar)
sentence = "peace is rising".split()

for tree in parser.parse(sentence):
    print (list(tree))

输出

[Tree('NP', [Tree('N', ['peace'])]), Tree('VP', [Tree('AUX', ['is']), Tree('V', ['rising'])])]

现在使用NP -> DT NP | N将限定词添加到NP中

import nltk

your_grammar = nltk.CFG.fromstring("""
S -> NP VP
VP -> AUX V 
V -> "rising"
AUX -> "is"
NP -> N | DT NP  
N -> "peace" | "price" 
DT -> "the"
""")

parser = nltk.RecursiveDescentParser(your_grammar)
sentence = "the price is rising".split()

for tree in parser.parse(sentence):
    print (list(tree))

输出

[Tree('NP', [Tree('DT', ['the']), Tree('NP', [Tree('N', ['price'])])]), Tree('VP', [Tree('AUX', ['is']), Tree('V', ['rising'])])]

最后，我们可以使用NP -> NP PP和PP -> P NP在NP中简单地添加PP结构

import nltk

your_grammar = nltk.CFG.fromstring("""
S -> NP VP
VP -> AUX V 
V -> "rising"
AUX -> "is"
NP -> N | DT NP | NP PP  
N -> "peace" | "price" 
DT -> "the"
PP -> P NP
P -> "of"
""")

parser = nltk.RecursiveDescentParser(your_grammar)
sentence = "the price of peace is rising".split()

for tree in parser.parse(sentence):
    print (list(tree))

这给了我们在顶部结果中最好的解析。

输出

[Tree('NP', [Tree('DT', ['the']), Tree('NP', [Tree('NP', [Tree('N', ['price'])]), Tree('PP', [Tree('P', ['of']), Tree('NP', [Tree('N', ['peace'])])])])]), Tree('VP', [Tree('AUX', ['is']), Tree('V', ['rising'])])]

但它也伴随着一些令人讨厌的递归循环错误，看起来像这样：

  File "/usr/local/lib/python3.5/site-packages/nltk/tree.py", line 158, in __getitem__
    return self[index[0]][index[1:]]
  File "/usr/local/lib/python3.5/site-packages/nltk/tree.py", line 156, in __getitem__
    return self[index[0]]
  File "/usr/local/lib/python3.5/site-packages/nltk/tree.py", line 150, in __getitem__
    if isinstance(index, (int, slice)):
RecursionError: maximum recursion depth exceeded in __instancecheck__

这是因为nltk.RecursiveDescentParser试图以递归方式查找解析，因为NP -> NP PP和PP -> P NP规则可以无限地重复。如果您想知道原因，请尝试在StackOverflow上以单独问题的形式提问;P

一个简单的解决方案是使用try-except

try:
    for tree in parser.parse(sentence):
        print (list(tree))
except RecursionError:
    exit()

但那是丑陋的！相反，您可以使用ChartParser

import nltk

your_grammar = nltk.CFG.fromstring("""
S -> NP VP
VP -> AUX V 
V -> "rising"
AUX -> "is"
NP -> N | DT NP | NP PP  
N -> "peace" | "price" 
DT -> "the"
PP -> P NP
P -> "of"
""")

parser = nltk.ChartParser(your_grammar)
sentence = "the price of peace is rising".split()

for tree in parser.parse(sentence):
    print (list(tree))

输出

[Tree('NP', [Tree('NP', [Tree('DT', ['the']), Tree('NP', [Tree('N', ['price'])])]), Tree('PP', [Tree('P', ['of']), Tree('NP', [Tree('N', ['peace'])])])]), Tree('VP', [Tree('AUX', ['is']), Tree('V', ['rising'])])]
[Tree('NP', [Tree('DT', ['the']), Tree('NP', [Tree('NP', [Tree('N', ['price'])]), Tree('PP', [Tree('P', ['of']), Tree('NP', [Tree('N', ['peace'])])])])]), Tree('VP', [Tree('AUX', ['is']), Tree('V', ['rising'])])]