Python NLTK: How to tag sentences with the simplified set of part-of-speech tags?

Question

Chapter 5 of the Python NLTK book gives this example of tagging words in a sentence:

>>> text = nltk.word_tokenize("And now for something completely different")
>>> nltk.pos_tag(text)
[('And', 'CC'), ('now', 'RB'), ('for', 'IN'), ('something', 'NN'), ('completely', 'RB'), ('different', 'JJ')]

nltk.pos_tag calls the default tagger, which uses a full set of tags. Later in the chapter a simplified set of tags is introduced.

How can I tag sentences with this simplified set of part-of-speech tags?

Also have I understood the tagger correctly, i.e. can I change the tag set that the tagger uses as I'm asking, or should I map the tags it returns on to the simplified set, or should I create a new tagger from a new, simply-tagged corpus?

Answer 1

To simplify tags from the default tagger, you can use nltk.tag.simplify.simplify_wsj_tag, like so:

>>> import nltk
>>> from nltk.tag.simplify import simplify_wsj_tag
>>> tagged_sent = nltk.pos_tag(tokens)
>>> simplified = [(word, simplify_wsj_tag(tag)) for word, tag in tagged_sent]

Answer 2

Updated, in case anyone runs across the same problem. NLTK has since upgraded to a "universal" tagset, source here. Once you've tagged your text, use map_tag to simplify the tags.

import nltk
from nltk.tag import pos_tag, map_tag

text = nltk.word_tokenize("And now for something completely different")
posTagged = pos_tag(text)
simplifiedTags = [(word, map_tag('en-ptb', 'universal', tag)) for word, tag in posTagged]
print(simplifiedTags)
# [('And', u'CONJ'), ('now', u'ADV'), ('for', u'ADP'), ('something', u'NOUN'), ('completely', u'ADV'), ('different', u'ADJ')]

Answer 3

You can simply set the tagset attribute to 'universal' in the pos_tag method.

In [39]: from nltk import word_tokenize, pos_tag
...: 
...: text = word_tokenize("Here is a simple way of doing this")
...: tags = pos_tag(text, tagset='universal')
...: print(tags)
...: 
[('Here', 'ADV'), ('is', 'VERB'), ('a', 'DET'), ('simple', 'ADJ'), ('way', 'NOUN'), ('of', 'ADP'), ('doing', 'VERB'), ('this', 'DET')]