Errors in edeps: double subjects/objects; and iobj vs. obj when only the recipient argument of a ditransitive verb is overt

UniversalDependencies / UD_English-EWT

English data

Creative Commons Attribution Share Alike 4.0 International

201 stars 43 forks source link

Errors in edeps: double subjects/objects; and iobj vs. obj when only the recipient argument of a ditransitive verb is overt #256

Open nschneid opened 3 years ago

nschneid commented 3 years ago

multiple E:obj dependents not explained by coordination: 51 EWT tokens, 7 GUM tokens
multiple subjects: 21 EWT tokens, 146 GUM tokens (some errors, others due to GUM's different policies regarding matrix copular constructions)
multiple E:iobjs: 1 erroneous GUM annotation
E:iobj but no obj/complement, and not a communication verb: 6 GUM matches, some erroneous

nschneid commented 1 year ago

Another double object example where the prepositional dative doesn't quite work all the time is cause:

It caused him pain.
?It caused pain to him.
"a medication intended to alleviate pain had, in fact, caused pain to so many"

amir-zeldes commented 1 year ago

As you just found, it's not so uncommon. More examples:

Bill Gates admits he 'caused pain' to Melinda when asked about cheating rumors (NY Daily News)
If anyone has caused pain, he has not caused pain to me (2 Corinthians 2:5)
The Veteran explained that during a difficult exercise he hurt his lower back which caused pain to "an already broken tailbone (va.gov)

I think we need to stick to syntactic criteria and say if it's a double object construction, our canonical analysis is iobj+obj, regardless of semantics. Then examples like 'cause' become less confusing IMO, even if their meanings are somewhat different from communication verbs or verbs of giving, the syntactic pattern applies.

nschneid commented 1 year ago

Group decision today: No universal prohibition on obj+ccomp—at least not for now—but we can go ahead and adopt Approach A within English, given that @amir-zeldes has the strongest opinion on this (and it's only slightly different from Approach B in practice).

nschneid commented 10 months ago

[x] "charge", at least, often has obj where iobj is correct. It seems I left "charge" out of queries in #55 (oops). The ones that need to be changed are all pronominal objects.