"quite" as predeterminer

[nschneid]

"quite" before a DET is tagged in most cases as PDT/DET and attached as det:predet. I see that this follows PTB guidelines, which list "quite" as a predeterminer:

However, this feels wrong to me. Unlike "all" and "both", "quite" can't be a regular DT (*I have quite shoe(s)). It behaves like a normal adverb before an adjective (quite tall), and seems awfully similar to "exactly":

• They're not quite/exactly identical.
• They're not quite/exactly the same.
• They're not quite/exactly your size.

The few EWT/GUM instances of "exactly" + DET are consistently ADV/advmod.

I don't mind keeping PDT on "quite" for backward compatibility but I feel ADV/advmod is the right way to go UD-wise.

[amir-zeldes]

I see what you're saying, but there are some differences, at least in the "quite a NOUN" construction. If you have "excatly" there, then it is a VP adverbial and you can do:

• That is exactly a dog
• That, exactly, is a dog
• That is a dog exactly
• That is exactly and positively a dog

But you can't do many of these with quite:

• ?That, quite, is a dog
• ??That is a dog quite
• ???That is quite and positively a dog

I think even if you accept any of the above, you would lose the exclamative sense, where "quite" belongs in the same family as "what a dog!" etc. So in sum I would be inclined to just leave this alone and let it aline with the established PDT tag.

[nschneid]

Other adverbs like "extremely" or "very" don't appear as modifiers in clause structure, either. And I believe "quite" has a wider distribution in British English than American English. https://simple.wiktionary.org/wiki/quite comments on this.

Exclamative sense: "Quite a dog!" doesn't work as a standalone utterance like "What a dog!", at least not for me. It has to be predicative ("That's quite a dog!"), which resembles "such" I think. But notably "what" and "such" have determiner uses that do not precede another determiner (and are not exclamative). And they don't modify adjectives (I don't think?): what tall, such tall.

"Quite" can also precede predeterminers: "Not quite all the members are on board" (maybe this should be advmod(all, quite) but in any case it shows that ADV rather than DET is typical for "quite").

[nschneid]

An example of modification in a clause: "I quite like the cake".

"Entirely" may be pretty close in distribution and meaning. "That is entirely/quite a bad idea"

[amir-zeldes]

Other adverbs like "extremely" or "very" don't appear as modifiers in clause structure, either.

Right, but we also don't normally see ?very a dog or ?extremely a dog, so that's in line with "quite" being special.

Exclamative sense: "Quite a dog!" doesn't work as a standalone utterance

I think it's pretty common... Examples from ENCOW:

• Quite a difference!
• Quite the contrary!
• Quite an honor for you!
• Quite the reverse.
• Quite a gap there, huh?
• Quite a remarkable ascent, I say.
• ...

Not quite all the members

I agree, this should be RB/advmod, it's not the same "quite" (it belongs with "not", not with "all" - you can't say "quite all the members")

An example of modification in a clause: "I quite like the cake".

This is also not that quite and should be RB/advmod IMO. The PDT one is just the one which is integrated into an NP and precedes the regular article.

[nschneid]

OK, "quite a..." can start a sentence (fragment) with exclamative meaning (not sure why "Quite a dog" sounds worse to me than "Quite a difference"....)...but isn't that true of many adverbs, like "Certainly"? ENCOW has things like

• Certainly a great looking , workmanlike , navigation station !
• Certainly a good start though !

In predicative contexts, "entirely a..." is quite common:

• But this is entirely a different question
• This was entirely a supernatural event
• So the 630 year project is not entirely a hypothetical

If "quite" has many clear ADV uses, and can occur before an article in much the same way as "certainly" or "entirely", I don't see why we need to say it transforms into a DET there.

[amir-zeldes]

Not that I'm in love with the PDT thing, but I think there are some differences there - the "entirely" examples are at the predication level IMO, not part of the NP like the "quite" we are talking about. You can do:

• A great looking , workmanlike , navigation station, certainly!
• A good start though, certainly!

But you can't normally do that with "quite" (though you can use it by itself as an utterance in response, meaning "that is quite so"). The difference is even more apparent in predications:

• But this is entirely a different question
• But this is a different question entirely
• ?? But this is a different question quite

I think these examples show that "quite" in the usage we are discussing is at the outermost border of a fuzzy category of originally adverbial and adjectival expressions which have snuck their way into the NP-internal domain and have idiosyncratic behavior in the PDT position, much like "such" (originally an adjective, but through bare plural usage, now a quasi determiner, with even more grammaticalized behavior e.g. in German)

[nschneid]

Hmm, true that the position of "entirely" is more flexible. Another interpretation is that "quite" is strictly a premodifier, like "very".

You bring up "such", and I had thought about "many" as an adjective that acquired determinerlike properties...yet we mostly keep the ADJ tag (#198). It hadn't occurred to me that they would be tagged DET only before an article. Isn't it strange to say that "many books" has no DETs but "many a book" has two?! We know there are contexts in which an adjective phrase can precede an article, e.g. "how good a horse" or "as good a horse".

Also, given that det:predet is a distinct function from det (unless we change that: #413) it seems unnecessary to be doubling the work at the tag level. I have argued for expanding DET in the past, but I thought the policy we arrived at has been to minimize UPOS ambiguity where possible. I take this to be why we have "the/DET many/ADJ books", "many/ADJ books", "many/ADJ of the books". While "many a book" is a different construction from these, perhaps warranting a different structure, the meaning of "many" is essentially the same.

[nschneid]

Vox populi:

That's 45 votes for adverb, 9 votes for determiner.

@erip points out that a word with a distribution similar to "quite" is "barely" (I would add "hardly" and "scarcely"):

• a barely/quite qualified teacher
• barely/quite a (qualified) teacher
• barely any teachers, not quite all teachers
• That is a different question entirely/??barely/??quite
• (preverbally, "quite" is more restricted to expressions like "quite like", but that doesn't really bear on how it should be tagged before an article IMO)

[amir-zeldes]

Hm, I don't put too much stock in Twitter surveys (maybe because I don't use it :) - but TBH I don't think the prompt "what's the POS in this example" is sufficient for getting a carefully considered answer, given the level of complexity and corner case examples we have in this discussion. I'm actually kind of surprised you got any 'determiner' responses with just that sentence...

Regarding 'barely', I think there are quite (heheh) a few differences:

• Not really used in the exclamative: "Quite a teacher!" vs. "?? Barely a teacher!"
• If you did get such uses, I suspect "barely" would be what I called a VP level or clause adverbial, not part of the NP like "quite"
• I think "quite" fills the same PDT slot as other PDTs, unlike barely. Notice you can combine barely with PDTs that clash with "quite":
  • Barely all the children
  • Quite the children!
  • ???Quite all the children

I agree with @erip that "not quite all" has quite as RB (it's "not quite" together, not "quite" as an NP determiner), so that's separate. I also agree with the preverbal point - those are all different, and we are just talking about the "quite an NP" type here.

I think ultimately this kind of 'quite' is on the border between the PDT class and its adverbial source, and it could probably fall on either side of the distinction. If we were having this discussion before any standards were made I wouldn't be too hard to convince to go with RB across the board, since a simple small tagset can't do justice to the subtleties either way. But since PTB guidelines already establish where it falls, I'd rather just stick to that and not introduce any more variants or dialects of English POS tag interpretations, esp. since we're not really redesigning the whole paradigm of PDT as a class. Lumping it with PDT "all" and "both" despite some differences is about as good as lumping it with "barely", and maintaining status quo tips the scales IMO.

[nschneid]

Hm, I don't put too much stock in Twitter surveys (maybe because I don't use it :) - but TBH I don't think the prompt "what's the POS in this example" is sufficient for getting a carefully considered answer, given the level of complexity and corner case examples we have in this discussion. I'm actually kind of surprised you got any 'determiner' responses with just that sentence...

I wouldn't treat any poll as determinative (ha) either, but a 5-1 ratio is interesting if we want to avoid treebanking policies that will surprise users...

Regarding 'barely', I think there are quite (heheh) a few differences:

• Not really used in the exclamative: "Quite a teacher!" vs. "?? Barely a teacher!"

I'm honestly not sure if the exclamativeness is something about the meaning of "quite", or idioms that "quite" tends to appear in, or what...but I'm not convinced it bears on the general syntactic category. I wouldn't expect most core predeterminers (all/both the books, half the pizza...) to be particularly exclamative.

• I think "quite" fills the same PDT slot as other PDTs, unlike barely. Notice you can combine barely with PDTs that clash with "quite":

  • Barely all the children
  • Quite the children!
  • ???Quite all the children

I agree with @erip that "not quite all" has quite as RB (it's "not quite" together, not "quite" as an NP determiner), so that's separate.

Again, there are probably some semantic and idiomaticity things going on here regarding which quantifiers and negations "quite" vs. "barely" are compatible with. Since "all" is a total quantifier it may be odd to intensify it, but negating it with "not quite" or negating its implicatures (?) with "barely" makes sense. "Barely", "scarcely", "hardly" have some negative meaning built in, unlike "quite", although "quite" does often occur in negative polarity contexts.

I don't think "quite" has to be completely interchangeable with another adverb to be considered an adverb. But there are quite a lot of similarities.

Also—Did you see my previous post? Whether "many a book" has two DETs is potentially relevant here.

[BrettRey]

CGEL isn't explicit in this section about Adv vs D, but much is the only D of the list.

Approximation: almost, nearly, practically, virtually, essentially, quite, much, rather

These modify superlatives, like the reinforcement modifiers, but in addition occur more readily than the latter with indefinite NPs: rather/∗ absolutely a good idea. Quite and much have both reinforcement and approximation uses: compare quite the worst response (reinforcement: “absolutely”) and quite a good idea (approximation: “fairly good”); much the best solution (reinforcement: “by far”) and much the same size (approximation: “virtually”). Nearly and much can themselves be modified by very: They had chosen very nearly/much the same material. This type also includes a number of idioms or fixed phrases: all but, more or less, to all intents and purposes, informal as good as. (CGEL p. 437)

[amir-zeldes]

CGEL isn't explicit in this section about Adv vs D, but much is the only D of the list

Thanks for checking - it looks like the distinction is a main focus there, and I guess they don't have a positionally defined category like PTB 'PDT' exactly

Did you see my previous post? Whether "many a book" has two DETs is potentially relevant here

PTB has 50-50 JJ vs. PDT for "many a" (two and two) and ON has the same split (10 cases, including the WSJ ones). But "quite/PDT" is very common and almost without exception.

Basically I think this 'quite' is on the border between an adverb and a PDT slot filler, sharing features of both. I could have been convinced to go with adverb pretty easily (they're both an ok-ish fit if these are the tags we have), but if we deprel it advmod then I think it should be upos ADV, but then I also think it should be xpos RB; and PTB/ON has a very clear precedent for PDT here (incl. in published guidelines), so I'm reluctant to change it without having any ability to re-do the LDC corpora, which are much bigger. Even EWT and GUM put together are less than a fifth of ON, so for training English POS taggers it's still very relevant, and I'm not keen on introducing inconsistencies, esp. when I don't think the original guidelines are totally off base.