Clausal dependents of adverbs

[LarsAhrenberg]

I have a problem with clausal dependents of adverbs such as

English: now that I've got your attention I'd like to ...
Swedish: nu när jag har er uppmärksamhet skulle jag vilja ... lit. now when I've got ...

In UD_English the analysis is ccomp(now,got) but this doesn't fit the definition of ccompas a core argument. For Swedish I'd prefer to use advcl: advcl(nu,har), but that would contradict the definition of advcl as a clause modifying a predicate. And acl wouldn't work either as that's for a dependent of a nominal. So, unless a fourth type of dependent clause is called for it seems one of the other three types needs a broader definition.

Lars

[amir-zeldes]

I feel like advcl is appropriate, otherwise maybe acl. Is it very different from:

That time when you went to the party

? I think this type of 'that' is more like 'when' than it is like 'which'.

[dan-zeman]

This problem is more general. I think we do not have guidelines (or at least have not stated them sufficently clearly and explicitly) for situations when the parent is neither predicate nor nominal. I have been struggling with obl vs. nmod in similar contexts.

I suspect that people lean towards the nominal vs. non-nominal division, i.e. advcl would be OK when the modified word is an adverb, acl would not.

[GPPassos]

Indeed, advcl documentation is restrictive to clauses modifying other clauses:

An adverbial clause modifier is a clause which modifies a verb or other predicate (adjective, etc.), as a modifier not as a core complement.

However, compare to advmod:

An adverbial modifier of a word is a (non-clausal) adverb or adverbial phrase that serves to modify a predicate or a modifier word.

advmod is very permissive, allowing modification of other modifier words*.

Shouldn't the definition for advcl be relaxed in order to be more parallel to advmod? This would solve the problem, the difference between the two being if the dependent is a clause or a modifier word.

*According to the function word modifiers section, it can even modify function words (although this isn't clearly stated in the advmod page)

[jnivre]

Yes, I think there are many definitions that need to reviewed. Unfortunately, work on the guidelines was stalled because of all the work that went into the CoNLL shared task and the porting of treebanks to v2. Improving the guidelines will be a top priority for the next period. More soon.

[sylvainkahane]

Same construction in French (Maintenant que vous êtes là, je voudrais vous dire que …).
For me it's clear that it's a relative clause:

I've got your attention now -> now that I've got your attention

Adverbs such as now are very close to temporal nouns and can be considered as temporal pronouns (Kahane S. (2010), Entre adverbes, noms et pronoms : le cas des modifieurs temporels, Actes de CMLF, La Nouvelle Orléans.):

now, today, Monday, the next day, …

Although the construction with now is quite idiomatic, it is also possible (at least in French) with the noun day and the next sentences are attested in French:

one day that I was in Paris, I …
the other day that I was coming back from the supermarket, I …

So I vote for acl:relcl with my two hands.

[nschneid]

I think "now that" + clause is a multiword expression/construction with a syntax similar to subordinator "so that" and meaning similar to subordinator "since". It can't be paraphrased as "currently that", nor is "yesterday that" or "soon that" possible. So I vote for fixed(now, that), mirroring fixed(so, that).

[jonorthwash]

I usually call adverbs with dependents adpositions. Or perhaps [non-attributive and non-core-argument] adpositions are just adverbs that take dependents.

(This comment may seem a bit trollesque, but I'm happy to discuss my reasoning, at least as it relates to certain cases or languages.)

[sylvainkahane]

@jonorthwash Yes ,adpositions are transitive adverbs (more or less said in Tesnière 1959 or Mel'cuk 1988). But some adverbs with a complement are not adpositions, if the complement is not direct (e.g. according (to sth)).

[jonorthwash]

But some adverbs with a complement are not adpositions, if the complement is not direct (e.g. according (to sth)).

I don't totally understand the reasoning here. "according to" seems like more of a canonical adposition than, say, "focussing on" (potentially not the best example).

[nschneid]

I don't have access to do a corpus search now, but isn't "according to" a multiword expression (fixed)?

…

On Sep 18, 2017 7:04 PM, "Jonathan Washington" ***@***.***> wrote: But some adverbs with a complement are not adpositions, if the complement is not direct *(e.g. according (to sth))*. I don't totally understand the reasoning here. "according to" seems like more of a canonical adposition than, say, "focussing on" (potentially not the best example). — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#488 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA8IryS69-h5E2_xiDKFi8Ufi0ACh4QVks5sjvbogaJpZM4PU8Lc> .

[jonorthwash]

I'm speaking more generally, not necessarily UD-specifically.

[sylvainkahane]

@nschneid Two answers:

1. Even if you consider according to as a MWE, you have to give a POS to according, which can never be ADP

2. that's really not the topic of this discussion, but I don't think that's a good idea to consider that to is part of MWE. This to is part of the sub-categorization (French rection) of according, exactly as to is part of the sub-categorization of talk in I talk to them. (Because you can say according to this and to that)

[amir-zeldes]

I'm not sure I follow what @jonorthwash is saying - is it the idea that every adverb that takes an argument is an adposition? I would think much like an adjective can take an argument, so can an adverb, and neither is necessarily an adposition as a result. I think it's easiest to look at deadjectival adverbs:

A similar idea to B is ...
amod(idea, similar)
obl(similar, B)

A works similarly to B
advmod(works, similarly)
obl(similarly, B)

In both cases I think it would be strange to claim that similar/similarly is an adposition, and I also think they should be analyzed as having dependents. But I might have misunderstood the issue.

[amir-zeldes]

@sylvainkahane about the current topic: I prefer advcl to acl, since I don't think it specifies a restrictive or non-restrictive relative clause:

"now that we know" != "that now which we know (and not other nows)"
"now that we know" != "a now, which we incidentally know"

I think advcl captures this paraphrase, which seems closer to intended reading:

"now, when we know, ..."

Do you feel that in this last example it's also acl? If not, is the presence of 'that' as the complementizer the criterion that you would use to choose acl over advcl, or does the meaning also play a role?

[nschneid]

I see @amir-zeldes's point that the ordinary meaning of "now" is somewhat transparent with "now that". OTOH, it seems to me that "now that" has a causal meaning mixed in that is not normally present with "now". Consider:

• I think we should get a cat now that we have a larger apartment.

"When?" does not seem like the most pertinent way to question the "now that..." part of the sentence, but rather, "Why?" (or "Why now?"). I think it is an idiosyncratic fact about "now that" that it can be paraphrased as "because" or "since".

I think this semantic idiosyncrasy supports what I argued earlier:

It can't be paraphrased as "currently that", nor is "yesterday that" or "soon that" possible. So I vote for fixed(now, that), mirroring fixed(so, that).

"now that" in English treebank

[sylvainkahane]

@amir-zeldes Yes the that-clause does a selection between all the possible now s.

now that you have finished your homework = 'many nows were possible and we are in a now in which you have finished your homework'

This is exactly what does a relative clause:

the day you didn't finished your homework = 'many days are possible and I consider the day in which you didn't finished your homework'

What is strange with this relative clause if the use of the relativizer that and the fact that it cannot be ellipsed. I'm far to be a native speaker but I suppose that now I have finished my homework is a bit strange.

In any case, even if the subcategorization frame of this now is unusual, I recommand not to treat that as fixed. It is part of the clause dependending on now. Cf. the discussion about to in #491. Of course there is here a phenomenon of lexicalization, but this concern the subcategorization of now. So it is possible to consider that this acl is become a ccomp. But it would be a bad idea to analyze the marker of this subcategorized clause as a part of a MWE.

[amir-zeldes]

@sylvainkahane yes, I'm also not happy with mwe, and I wouldn't think this is ccomp for the clause, since you can use now equally without the clause. The 'that' is not mandatory, you can omit it as in:

So now we have an idea about what love is , let 's think about things that we love . (UKWAC)

What I find more curious is that you can't use 'which':

* Now which you've finished your homework...

I'm still not sure I see the diagnostic from distinguishing the subordinate clause as acl rather than advcl - how is it different from "now when you've finished your homework" or 'once' etc. I guess that can also be paraphrased as selecting from alternatives: "that particular now, when ...". In a real adnominal clause expanding 'now', I would expect the following to be possible:

"I experience a now that/which/*when is more vibrant than any other".

In this kind of example it's clear that 'now' is acting as the head of an NP, and a normal acl is possible. For "now that" it feels more adverbial to me.