search

UniversalDependencies / docs

Universal Dependencies online documentation
http://universaldependencies.org/
Apache License 2.0
274 stars 249 forks source link

English ccomp/acl #93

Closed mcdm closed 10 years ago

mcdm commented 10 years ago

@ngiordani @manning I am going through other examples that we had before in SD/USD, to see how they should be analyzed with the new guidelines. Do you agree with the following?

CCOMP

They heard about you missing classes
ccomp(heard, missing)
mark(missing, about)
dobj(missing, classes)
nsubj(missing, you)
I am certain that he did it
ccomp(certain, did)
mark(did, that)

What about "the fact that": ACL ?

Right now, in "ccomp" for English we say (that was in our manual for SD): Clausal complements for nouns are limited to complement clauses with a subset of nouns like fact or report. We analyze them the same (parallel to the analysis of this class as "content clauses" in Huddleston and Pullum 2002). Such clausal complements are usually finite (though there are occasional remnant English subjunctives).

But this contradicts the new guidelines for annotating clauses. So now that should be "acl"?

I admire the fact that you are honest
acl(fact, honest)
mark(honest, that)
cop(honest, are)
nsubj(honest, you)
ngiordani commented 10 years ago

I agree with ccomp for

I am certain that he did it

but I think we should probably use advcl for (heard, missing), because it seems like a noncore argument to me...

N.

On Wed, Oct 8, 2014 at 9:05 AM, mcdm notifications@github.com wrote:

@ngiordani https://github.com/ngiordani @manning https://github.com/manning I am going through other examples that we had before in SD/USD, to see how they should be analyzed with the new guidelines. Do you agree with the following?

CCOMP

They heard about you missing classes ccomp(heard, missing) mark(missing, about) dobj(missing, classes) nsubj(missing, you)

I am certain that he did it ccomp(certain, did) mark(did, that)

What about "the fact that": ACL ?

Right now, in "ccomp" for English we say (that was in our manual for SD): Clausal complements for nouns are limited to complement clauses with a subset of nouns like fact or report. We analyze them the same (parallel to the analysis of this class as "content clauses" in Huddleston and Pullum 2002). Such clausal complements are usually finite (though there are occasional remnant English subjunctives).

But this contradicts the new guidelines for annotating clauses. So now that should be "acl"?

I admire the fact that you are honest acl(fact, honest) mark(honest, that) cop(honest, are) nsubj(honest, you)

— Reply to this email directly or view it on GitHub https://github.com/UniversalDependencies/docs/issues/93.

ngiordani commented 10 years ago

Closing this, as we settled it via email. We'll go with

They heard about you missing classes. advcl(heard, missing)

to keep the parallel with

They heard about your bad grades. nmod(heard, grades)