case/det/mark attachment in unusual environment

[nschneid]

Need to review:

• [x] case/det: https://universal.grew.fr/?custom=660c06384ab10
• [x] mark: https://universal.grew.fr/?custom=660c06c0439f9

[nschneid]

"about the same noise level" is

but it seems that elsewhere, approximators "about", "approximately", "almost" followed by "the" attach to a head noun unless the adjective "same" (or "last") is promoted to head of the nominal: https://universal.grew.fr/?custom=661af51245862

So I'll change both "about" and "the" to attach to "level".

[nschneid]

That leaves this weird case—"terms substantially the same as what's in the draft":

"same" is analyzed as a adjective post-modifier—but then why does it have its own determiner? The determiner cannot be separated from "same": *We negotiated the terms same as what's in the draft (non-paratactic interpretation)

Is it like a size/extent modifier NP: "dogs [the size of elephants]"? That would suggest a noun head could be inserted, but: *terms the same terms/policies as what's in the draft

@amir-zeldes thoughts?

[nschneid]

A few of the mark ones are part of the larger clausal appos problem: UniversalDependencies/docs#1024

[amir-zeldes]

That would suggest a noun head could be inserted, but: *terms the same terms/policies as what's in the draft

Not sure follow this - you also can't say "dogs the size dogs of elephants", so in that respect they are similar. I guess nmod:npmod would be OK?

[nschneid]

Well "size" is a noun so it is expected to head a nominal. But "same" is an adjective, so I was wondering whether we take "same" as promoted from "same N".

[amir-zeldes]

Oh yeah, I think that's what I was trying to say with nmod:npmod (so treat it like an NP)

[nschneid]

But if it was promoted, we would expect to be able to demote it by inserting a head noun, right? I guess you could say "terms substantially the same as X" is short for "terms that are substantially the same (terms) as X". So yeah, maybe nmod:npmod works. I could see annotators being confused about why it's a different deprel from "terms equal to X" (that would be amod because there's no determiner before the adjective).

Or...is the relative clause paraphrase a red herring? That would use amod after all.

[amir-zeldes]

I think nmod:npmod is the best option here, and defensible/distinguishable if annotators ask, since we have the "the", which is unusual and does seem like a pretty strong indication that amod is not quite right.

[nschneid]

closing this as clausal appos is in another issue