How to use pos_tag in NLTK?

SSBakh picture SSBakh · Nov 27, 2017 · Viewed 22.5k times · Source

So I was trying to tag a bunch of words in a list (POS tagging to be exact) like so:

pos = [nltk.pos_tag(i,tagset='universal') for i in lw]

where lw is a list of words (it's really long or I would have posted it but it's like [['hello'],['world']] (aka a list of lists which each list containing one word) but when I try and run it I get:

Traceback (most recent call last):
  File "<pyshell#183>", line 1, in <module>
    pos = [nltk.pos_tag(i,tagset='universal') for i in lw]
  File "<pyshell#183>", line 1, in <listcomp>
    pos = [nltk.pos_tag(i,tagset='universal') for i in lw]
  File "C:\Users\my system\AppData\Local\Programs\Python\Python35\lib\site-packages\nltk\tag\__init__.py", line 134, in pos_tag
    return _pos_tag(tokens, tagset, tagger)
  File "C:\Users\my system\AppData\Local\Programs\Python\Python35\lib\site-packages\nltk\tag\__init__.py", line 102, in _pos_tag
    tagged_tokens = tagger.tag(tokens)
  File "C:\Users\my system\AppData\Local\Programs\Python\Python35\lib\site-packages\nltk\tag\perceptron.py", line 152, in tag
    context = self.START + [self.normalize(w) for w in tokens] + self.END
  File "C:\Users\my system\AppData\Local\Programs\Python\Python35\lib\site-packages\nltk\tag\perceptron.py", line 152, in <listcomp>
    context = self.START + [self.normalize(w) for w in tokens] + self.END
  File "C:\Users\my system\AppData\Local\Programs\Python\Python35\lib\site-packages\nltk\tag\perceptron.py", line 240, in normalize
    elif word[0].isdigit():
IndexError: string index out of range

Can someone tell me why and how I get this error and how to fix it? Many thanks.

Answer

alvas picture alvas · Nov 28, 2017

Firstly, use human-readable variable names, it helps =)

Next, pos_tag input is a list of string. So it's

>>> from nltk import pos_tag
>>> sentences = [ ['hello', 'world'], ['good', 'morning'] ]
>>> [pos_tag(sent) for sent in sentences]
[[('hello', 'NN'), ('world', 'NN')], [('good', 'JJ'), ('morning', 'NN')]]

Also, if you have the input as raw strings, you can use word_tokenize before pos_tag:

>>> from nltk import pos_tag, word_tokenize
>>> a_sentence = 'hello world'
>>> word_tokenize(a_sentence)
['hello', 'world']
>>> pos_tag(word_tokenize(a_sentence))
[('hello', 'NN'), ('world', 'NN')]

>>> two_sentences = ['hello world', 'good morning']
>>> [word_tokenize(sent) for sent in two_sentences]
[['hello', 'world'], ['good', 'morning']]
>>> [pos_tag(word_tokenize(sent)) for sent in two_sentences]
[[('hello', 'NN'), ('world', 'NN')], [('good', 'JJ'), ('morning', 'NN')]]

And you have the sentences in a paragraph, you can use sent_tokenize to split the sentence up.

>>> from nltk import sent_tokenize, word_tokenize, pos_tag
>>> text = "Hello world. Good morning."
>>> sent_tokenize(text)
['Hello world.', 'Good morning.']
>>> [word_tokenize(sent) for sent in sent_tokenize(text)]
[['Hello', 'world', '.'], ['Good', 'morning', '.']]
>>> [pos_tag(word_tokenize(sent)) for sent in sent_tokenize(text)]
[[('Hello', 'NNP'), ('world', 'NN'), ('.', '.')], [('Good', 'JJ'), ('morning', 'NN'), ('.', '.')]]

See also: How to do POS tagging using the NLTK POS tagger in Python?