python - Extract Word from Synset using Wordnet in NLTK 3.0

14
votes

Some time ago, someone on SO asked how to retrieve a list of words for a given synset using NLTK's wordnet wrapper. Here is one of the suggested responses:

for synset in wn.synsets('dog'):
    print synset.lemmas[0].name

Running this code with NLTK 3.0 yields TypeError: 'instancemethod' object is not subscriptable.

I tried each of the previously-proposed solutions (each of the solutions described on the page linked above), but each throws an error. I therefore wanted to ask: Is it possible to print the words for a list of synsets with NLTK 3.0? I would be thankful for any advice others can offer on this question.

3
pythonnlpnltkwordnet
Is there a reason for using the latest NLTK versus more stable (albeit earlier) version? - user3898238
I'm working in multilingual contexts, and I thought the Open Multilingual Wordnet was only implemented in 3.0. If it was available in earlier iterations, though, I'm all ears! - duhaime
It looks like OMW has been available via NLTK for some time now. I'll try reverting to an earlier and more stable release... - duhaime
Could you show where your attempts fail, e.g. what you tried and the errors/complaints? - user3898238
The specific error message and a MWE would be helpful. - Dan

3 Answers

15
votes

WordNet works fine in NLTK 3.0. You are just accessing the lemmas (and names) in the wrong way. Try this instead:

>>> import nltk
>>> nltk.__version__
'3.0.0'
>>> from nltk.corpus import wordnet as wn
>>> for synset in wn.synsets('dog'):
    for lemma in synset.lemmas():
        print lemma.name()


dog
domestic_dog
Canis_familiaris
frump
dog
dog
cad
bounder
blackguard
...

synset.lemmas is a method and does not have a __getitem__() method (and so is not subscriptable).

10
votes

You can also go directly to the lemma names with lemma_names(): 

>>> wordnet.synset('dog.n.1').lemma_names()
['dog', 'domestic_dog', 'Canis_familiaris']

And it works for multiple languages

>>>> wordnet.synset('dog.n.1').lemma_names(lang='jpn')
['イヌ', 'ドッグ', '洋犬', '犬', '飼犬', '飼い犬']
7
votes

Use:

wn.synset('dog.n.1').name()

instead of:

wn.synset('dog.n.1').name

because NLTK changed Synset properties to get functions instead. see https://github.com/nltk/nltk/commit/ba8ab7e23ea2b8d61029484098fd62d5986acd9c

This is a good list of changes to NLTK's API to suit py3.x: https://github.com/nltk/nltk/wiki/Porting-your-code-to-NLTK-3.0