own-pt/openWordnet-PT

topology of PWN

vcvpaiva opened this issue · 3 comments

vcvpaiva commented

Could anyone tell me how many of the 117659 synsets have glosses? not all do

Can we add to the repo somewhere the corpus of glosses, inspectable?
https://wordnet.princeton.edu/glosstag.shtml

fcbr commented

@vcvpaiva AFAIK all PWN synsets have glosses. For example, if we use the Prolog output of PWN, and removing the duplicates in the Prolog generated we have 117659 entries.

$ cd prolog
$ cat wn_g.pl | awk -F, '{print $1}'| sort | uniq -c | wc
117659

Also, if we look at the tagged glosses, it seems that all of them have tagged glosses too:

$ cd glosstag/standoff
$ cat index.byid.tab | awk -F'$\t' '{print $1'} | sort | uniq | wc
117659
vcvpaiva commented

thanks @fcbr!
this is odd, as I am sure many times I have had the impression not having a gloss.
maybe it's when it's a single word like

05893261-n sine_qua_non, essential_condition | sine qua non
(a prerequisite)

what is a tagged gloss, please?

and questions on the topology of PWN:

  1. how many synsets s go directly to Entity? do all synsets go to Entity?

  2. how many have two hops?

  3. how many have a long hierarchy like kitty<domestic_cat <cat<feline<carnivore< placental_mammal < mammal<vertebrate<chordate<animal<organism<living_thing<entity?

  4. I seem to remember that you were calculating isolated nodes vs hierarchies? where is that data now?

arademaker commented

Yes, we do have glosses with only 1-2 words. The tagged gloss corpus is not complete; not all glosses are entirely tagged, I talked with Christiane Fellbaum about it. Actually, this is an excellent work still waiting to be done.

corpus of glosses = tagged corpus