Partitive headed by an ADJ quantifier ("many", "much", "few", etc.) should be nmod not obl

[nschneid]

Inconsistent in GUM: https://universal.grew.fr/?custom=69075d1675776

It is consistently nmod in EWT except for one error ("more and more of").

[amir-zeldes]

Those are a lot of different lemmas in several constructions... Are you saying you'd want all PP dependents of all instances of /many|most|more|much|little|few/ to be nmod, or just when there's "of"? I could see making them all obl categorically based on pos, but some of them don't looks like nmod is right to me. This kind seems adjectival:

• Some people would consider it more of a personal outlook than a learning practice

And there are borderline ones with advmods like this:

• one of very few of its kind

If few is nominalized, I'm not sure how to explain "very" here...

[nschneid]

I was thinking all "of" ones where the of-PP denotes the superset/supercategory.

Predicative "more of a(n) X" is a metalinguistic idiom, I guess; it's not immediately obvious what the syntax should be. There is one EWT token.

• one of very few of its kind

If few is nominalized, I'm not sure how to explain "very" here...

I don't think it's nominalized; I think it's promoted in the absence of a following noun. So it can still have an advmod.

[amir-zeldes]

I think it's promoted in the absence of a following noun

I'm not 100% sure that's right... I'm not saying I'm not open to it, but I think I need a little more convincing. For example, if we have:

• We shouldn't make so much of it.

The "so" is an advmod, so I think locally it's clear much is an adjective. If this refers to reducing plastic production, then "it" refers back to plastic. In the elliptical noun reading, I could say:

• We shouldn't make so much plastic

But now the whole "of it" is gone... I don't think it's normal to say:

• * We shouldn't make so much plastic of it

So that seems to suggest "much" isn't an elliptical version of "much plastic", at least not in cases where the partitive is there - it's a properly adjectival head projecting an NP. So I agree the whole thing ends up being an NP, and maybe that alone is already reason enough to go with nmod, but I don't quite get the promotion reading because of this discrepancy.

[nschneid]

"make so much of it" is an idiom. There are variants like "make so much hay of it" or "make so much fuss about it".

[amir-zeldes]

No, that's not it - I should have given more context. The actual text reads:

• ... the only way to stop ocean plastic is to stop plastic ... The better solution would be to not make so much of it to begin with"

So "it" really means plastic here.

[nschneid]

Oh OK. So this is essentially: We shouldn't make so much plastic / so much of the plastic.

It's a lot like "some" or "any" (We don't make any plastic / any of the plastic). We don't say "any plastic of it", either; that's just how partitive constructions work (sometimes they are analyzed as fused-head NPs).

[amir-zeldes]

We don't say "any plastic of it"

Exactly - if it were just promotion due to ellipsis, then "much" (or any) should stand for "much plastic" (or any plastic), and that doesn't work for partitives.

that's just how partitive constructions work

Maybe so, but PP modifiers of adjectives are normally obl, so we would need to clarify what is the criterion that licenses nmod here. Semantically I get it - they are referring expressions, and syntactically I get it from the outside perspective - the result is an NP. But internally, it looks like the of-PP is just a dependent of an adjectival head, so it makes me unsure if this really shouldn't be obl. I'm open to making them nmod, but I'd like to understand what the criterion is. I think @dan-zeman often has strong intuitions about nmod vs. obl - do you have an opinion?

[nschneid]

"I didn't use much (plastic) of the plastic you sent me" is OK though, if a bit redundant.

[dan-zeman]

Does the validator flag these instances? I think it shouldn't; if I recall correctly what I wrote, obl-should-be-nmod is only tried when the parent is NOUN|PROPN|PRON.

Nevertheless, I'm unsure which of the two deprels we should require in these constructions. If it modifies a nominal, it should be nmod. Here it seems to be the case, as indicated by the relation incoming to the parent (I did not inspect all of them; but in the first example it is obj). On the other hand, we use obl for modifiers of adjectives. But my interpretation is that if the adjective happens to head a nominal, we only use obl if the dependent modifies the adjective alone, not the whole nominal (the latter would mean there is ellipsis). I'm inclined to think that in partitives like many of us, the dependent modifies just the adjective and is thus obl. But it's definitely not a strong intuition :-) Then again, it is unclear why we should be allowed to make the whole thing an obj of a verb, if the thing is just an adjective and not a nominal

[nschneid]

I don't think these are flagged by the validator.

Note that with "all", "many", etc., you can add a head noun like "instances" or "members":

• all (instances) of the books
• many (instances) of the books

That doesn't work so well with mass nouns (no equivalent of 'instances'—a lexical gap?):

• ?much material of the plastic

But again, if the definite NP in the second part of the partitive is more elaborate, it can work:

• much (plastic) of the plastic you sent me

For me, the partitive construction is fundamentally a nominal relation, not an oblique. If we really need to emphasize the DET-like behavior of "much" and "many" (as opposed to "plentiful", "numerous") with respect to partitives, we could consider ExtPos=DET. But I think the promotion interpretation is close enough to explaining the partitive construction generally, even if not every instance is a good fit for an inserted noun. (Even textbook ADJ promotion cases like "the poor" are not merely a matter of inserting a noun: the better paraphrase is "poor people" not "the poor people".)