Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Tagged as nominal, should actually be VERB? #111

Open
nschneid opened this issue Feb 1, 2022 · 4 comments
Open

Tagged as nominal, should actually be VERB? #111

nschneid opened this issue Feb 1, 2022 · 4 comments

Comments

@nschneid
Copy link
Contributor

nschneid commented Feb 1, 2022

http://match.grew.fr/?corpus=UD_English-GUM@dev&custom=61f88fb74e07f

@amir-zeldes
Copy link
Owner

This is a tricky one... We don't manually tag upos, and for NNPs that are VERB (or ADJ) we rely on some obvious deprels in the conversion (e.g. NNP+amod -> VERB if .*(ed|ing) and lemma!=form, else ADJ)

The cases this query identifies are heads, so it's not easy to find all of them. I agree the ones that have obj must be verbs (and same for .*:pass and a few other things), but even that won't tell us the right morphology. Consider some hypothetical movies called:

  • I Fooled/NNP Destiny
  • I have Fooled/NNP Destiny
  • Fooled

For 1-2 we can tell they are verbs because they have obj; but I can't tell whether they are VerbForm=Part or Fin. In the last case I'm not sure I can tell anything - is it a verb? Finite? An adjective? I could write some rules to catch maybe 70% of cases, but I'm not sure if that's actually better than leaving it (at least then it's consistent)

@nschneid
Copy link
Contributor Author

nschneid commented Feb 3, 2022

Not consistent with other English corpora, though. I would err on the side of tagging it VERB if it can be interpreted as a verb.

@amir-zeldes
Copy link
Owner

Right, I'm not saying I want it to be inconsistent with other corpora, I'm just saying I have no means of doing it automatically in a reliable way, and currently not enough resources to manually go over all NNPs in the corpus. I will leave this issue open in case we have someone who can do it in the future. I'll add the help-wanted tag, but anyone thinking of helping with this should talk with me first, since it would need to be done pre-conllu in _build/src/

@nschneid
Copy link
Contributor Author

nschneid commented Feb 7, 2022

Note for whoever takes this on: WordNet may be helpful here.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants