Nltk: Function to access the Synset object using sense key
I might have missed it but is there a function to access the Synset object from the NLTK wordnet interface from the sense key?
If there isn't could we expose a function that can achieve that in nltk.corpus.wordnet? E.g. https://stackoverflow.com/questions/48170666/how-to-get-the-gloss-given-sense-key-using-nltk-wordnet/
Ideally, it would be good to have functionalities to access Synset objects using:
offset-pos, e.g.1433493-a->Synset('long.a.02')sense_key, e.g.long%3:00:02::->Synset('long.a.02')
Currently, we have the synset_from_pos_and_offset() for (1).
There's another function (_synset_from_pos_and_line) that reads the following line to return Synset('long.a.02'):
01433493 00 a 01 long 1 016 = 05129201 n 0000 + 05133287 n 0101 ! 01436003 a 0101 & 01434007 a 0000 & 01434218 a 0000 & 01434530 a 0000 & 01434717 a 0000 & 01434841 a 0000 & 01434966 a 0000 & 01435060 a 0000 & 01435189 a 0000 & 01435290 a 0000 & 01435399 a 0000 & 01435507 a 0000 & 01435675 a 0000 & 01435891 a 0000 | primarily spatial sense; of relatively great or greater than average spatial extension or extension as specified; "a long road"; "a long distance"; "contained many long words"; "ten miles long"
but it's not the sense key.
alvations
All 4 comments
I'd like to work on this!
craaaa
on 3 Feb 2018
@craaaa Sorry for the late reply, was away for a while.
Feel free to work on it and create a PR afterwards.
P/S: Don't worry about breaking anything, there'll be checks and reviews before we merge the code.
alvations
on 15 Feb 2018
Implemented the function suggested in the stackoverflow, but it didn't seem to map to the correct senses -- for instance, synset_from_sense_key('afraid%3:00:00::') returned afraid.a.04 instead of afraid.a.01. This problem extends to other POS as well. (Sense keys were obtained from Wordnet's online interface)
Instead, using the method shown in the SemCor documentation appears to map correctly -- there is currently a lemma_from_key(key) function that seems to take in something similar to a sense key. However, lemma_from key(key) doesn't support adjective satellites (e.g. afraid%3:00:02:concerned:00). I can definitely implement a wrapper around lemma_from_key(key) to fix this and return a Synset.
craaaa
on 15 Feb 2018
@craaaa. I think there is still an issue with adjective satellites. I tried to use the synset_from_sense_key function and here is the error:
File "/home/izorar/anaconda3/lib/python3.7/site-packages/nltk/corpus/reader/wordnet.py", line 1356, in synset raise WordNetError(message % lemma) WordNetError: adjective satellite requested but only plain adjective found for lemma 'first'
Any idea on how to fix the error?
Izorar
on 27 Oct 2019
Related issues
chaseireland
·
3Comments
mwess
·
5Comments
BLKSerene
·
4Comments
stevenbird
·
4Comments
jeryini
·
5Comments
Most helpful comment
Implemented the function suggested in the stackoverflow, but it didn't seem to map to the correct senses -- for instance,
synset_from_sense_key('afraid%3:00:00::')returnedafraid.a.04instead ofafraid.a.01. This problem extends to other POS as well. (Sense keys were obtained from Wordnet's online interface)Instead, using the method shown in the SemCor documentation appears to map correctly -- there is currently a
lemma_from_key(key)function that seems to take in something similar to a sense key. However,lemma_from key(key)doesn't support adjective satellites (e.g.afraid%3:00:02:concerned:00). I can definitely implement a wrapper aroundlemma_from_key(key)to fix this and return aSynset.