amir-zeldes / gum

Repository for the Georgetown University Multilayer Corpus (GUM)
https://gucorpling.org/gum/
Other
89 stars 50 forks source link

obj should be iobj #146

Closed nschneid closed 1 year ago

nschneid commented 1 year ago

With pronouns: https://universal.grew.fr/?custom=644ebd584cb1a But in GUM this should be extended to all animate nouns.

Tricky case: "Thank God that..." (and similar) means that "thank" can license an iobj+ccomp. Do we want to say that "thank" objects should always be iobj, or treat this as an exceptional idiom?

nschneid commented 1 year ago

related to UniversalDependencies/UD_English-EWT#55

nschneid commented 1 year ago

Also a bit borderline: "believe" and "trust". But these can have an object + ccomp: "Trust me that it will happen." COCA: "He really believed me that I was going to do this for him."

amir-zeldes commented 1 year ago

Thank God that

That is admittedly very marginal. We have no cases of thank+ccomp in GUM at all, and that's 200K tokens from very different genres... It's also not a typical ccomp since it's not the contents of the thanks, it's the reason (equiv. to "thank X for Y"). But if you want iobj here I guess I can live with it (incidentally, danken takes dative in German, suggesting there is some historical evidence it is somehow not a proper accusative object)

Also a bit borderline: "believe" and "trust"

I would be more OK with these taking iobj (believe X that Y sounds quite normal, and it is the content of the belief, so it's a pretty canonical case IMO)

nschneid commented 1 year ago

OK.

amir-zeldes commented 1 year ago

iobj in general for "believe" and "trust" (a source/authority)

Just implemented in GUM, thanks!

a Divine Exception of iobj

😂