Nominal measurement phrase + PP: "5 days after the election"

[nschneid]

I've noticed this is treated inconsistently across the few treebanks I have checked. Two issues are (a) headedness and (b) deprel.

Measurement phrase as head

PP attaches as nmod: This seems to be the convention in Czech-PDTC:

Google Translate: 'and three days after his inauguration in January, he reaffirmed his "firm support for our cause" at an anti-abortion rally.'

What is the reason behind making the measurement phrase the head? Can the PP be omitted?

Measurement phrase as modifier

As far as I can tell, this is the convention in English, Swedish, and French, though the deprel differs:

• Measurement phrase as nmod: An example in French-Sequoia

  

• Measurement phrase as nmod:unmarked: This is the convention in EWT

  

• Measurement phrase as obl: A couple of examples in Swedish-Talbanken

  

• Measurement phrase as obl:unmarked: This is the convention in English-GUM

  

The :unmarked convention may be somewhat English-specific, so I am mainly worried about nmod vs. obl. Which one is better?

(N.B. There is also the related construction where the second part is a subordinate clause, e.g. "5 days after the election took place". If "5 days" is a modifier, then it is clearly obl:unmarked here.)

[dan-zeman]

(Czech) What is the reason behind making the measurement phrase the head? Can the PP be omitted?

In the particular example shown here (with a perfective verb), the prepositional phrase po lednové inauguraci "after the inauguration in January" cannot be omitted. If the verb were imperfective, the PP could be omitted. On the other hand, the measurement tři dny "three days" can be omitted in this example but it could not be omitted with a diffferent preposition: tři dny od lednové inaugurace, unless the verb was imperfective.

Also, all the above holds under the assumption that the whole expression is used as a temporal adjunct (as it is in this example). It could be also a core argument, and then the "days" would be the head without any doubt (tři dny po inauguraci byly nejtěžší "the three days after the inauguration were the hardest").

Whether/which of the above were actually used as the criteria for headedness in PDT-C is a different question. I was not involved in the decision, perhaps @mikulova knows more.

Corresponding PDT-C query here.

[nschneid]

Also, all the above holds under the assumption that the whole expression is used as a temporal adjunct (as it is in this example). It could be also a core argument, and then the "days" would be the head without any doubt (tři dny po inauguraci byly nejtěžší "the three days after the inauguration were the hardest").

Sure—for this issue I'm interested in the cases where the "days" phrase measures an interval between the two events, rather than referring to a particular set of days. The same construction (at least in English) also works for spatial measures: "They live 5 miles down the road/past the school".

[dan-zeman]

The same construction (at least in English) also works for spatial measures: "They live 5 miles down the road/past the school".

In the Czech equivalent of this (bydlí 5 mil od školy), neither 5 mil nor od školy can be omitted.

[mikulova]

Sorry, I'm on vacation. Currently on my way from Brazil to Europe.   These constructions are a subject of discussion in Czech linguistics.  I'll write more on Wednesday at the latest. Marie Mikulová   

…

-----Original Message----- From: Nathan ***@***.***> To: UniversalDependencies ***@***.***> Cc: mikulova ***@***.***>; Mention ***@***.***> Date: Sunday, 16 November 2025 3:40 PM -03 Subject: Re: [UniversalDependencies/docs] Nominal measurement phrase + PP: "5 days after the election" (Issue #1175) nschneid left a comment (UniversalDependencies/docs#1175) Also, all the above holds under the assumption that the whole expression is used as a temporal adjunct (as it is in this example). It could be also a core argument, and then the "days" would be the head without any doubt (tři dny po inauguraci byly nejtěžší "the three days after the inauguration were the hardest"). Sure—for this issue I'm interested in the cases where the "days" phrase measures an interval between the two events, rather than referring to a particular set of days. The same construction (at least in English) also works for spatial measures like "5 miles". — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: ***@***.***>

[sylvainkahane]

It is very clear that the measurement-phrase is not a noun modifier. Even if it attached to a noun because UD constraints do not allow us to attach it to the preposition, it is validated by the preposition. It seems that according to UD constraints obl or obl:unmarked is the required solution.
Note also that the measurement-phrase can depend on a (pronominal) adverb (she came five days before) and in this case nmod would be absurd.

[Stormur]

At a first look I would say that these have all to be treated as arguments of the predicate, so as separate obls. They are not really subordinated one to the other, but interconnected through the predicate, as Dan discussed with regard to aspect.

There are probably also some preferences with regard to specific ordering, so e.g. the more generic time argument comes before the more specific ones. The same when you say "This Monday at the beginning of the class I explained that..."

[Stormur]

Note also that the measurement-phrase can depend on a (pronominal) adverb (she came five days before) and in this case nmod would be absurd.

It would not, actually, because even if we keep calling them "adverbs" some of these elements are clearly nominals. Or anyway, it can be that an adpositional element can stay for the whole phrase.

[nschneid]

It is very clear that the measurement-phrase is not a noun modifier. Even if it attached to a noun because UD constraints do not allow us to attach it to the preposition, it is validated by the preposition. It seems that according to UD constraints obl or obl:unmarked is the required solution. Note also that the measurement-phrase can depend on a (pronominal) adverb (she came five days before) and in this case nmod would be absurd.

Yes, it seems that (for English) the measurement phrase is possible before "before":

• Before the elections, Pakistan closed the border crossing.
• Several days before the elections, Pakistan closed the border crossing.
• Several days before, Pakistan closed the border crossing.
• *Several days, Pakistan closed the border crossing.
• *Several days, Pakistan closed the border crossing before the elections.

But in defense of the English-EWT policy (nmod:unmarked): UD does not formally recognize PPs or allow contentful modification of ADPs. So I think we have a nominal that modifies another nominal, but outside its "core", as signaled by the intervening case dependent.

We actually recognize a class of predeterminers (e.g. "all" in "all the books") and attach them as det:predet because they are outside the core nominal. I have some quibbles with the details of that policy. But if we are to be consistent, surely the measurement phrase is even farther from the head—separated not only by a determiner but also a preposition. Could we call it nmod:precase???

Note also that the measurement-phrase can depend on a (pronominal) adverb (she came five days before) and in this case nmod would be absurd.

It would not, actually, because even if we keep calling them "adverbs" some of these elements are clearly nominals. Or anyway, it can be that an adpositional element can stay for the whole phrase.

Others have argued that "before" should have a consistent part of speech, whether transitive or intransitive. But UD treats them differently (ADP vs. ADV). I think any dependent of an ADV has to be obl rather than nmod.