a kos vayn / ein Glas Wein / a glass of wine

[shoowadoo]

In German treebanks, constructions such as eine Menge Formulare 'a bunch of forms' use the relation appos. I imagine a similar analysis would be applied to ein Glas Wein 'a glass of wine.' In these constructions, there is no preposition between the two nouns.

In Yiddish, there is a similar phenomenon, e.g., a kos vayn 'a glass of wine'.

However, I'm not sure whether these elements can be swapped and whether they fulfill the criteria for using appos. nmod:unmarked is used in English for things like 'a couple cookies' and the like. I'm wondering if this is a better option than using appos or maybe even just plain nmod?

[dan-zeman]

I would use plain nmod. I didn't realize the German treebanks used appos for this construction, it does not feel right.

[amir-zeldes]

I also don't think it's appos. If the :unmarked subtype is used, then it would fit here, but either way nmod sounds like the right major type.

[Stormur]

nmod is the right one.

The only other question here is: what is the nmod of what? Because, also for agreement phenomena, we see that eine Menge, ein Glas, ... go towards classifiers/determiners.

[LeonieWeissweiler]

This feels very similar to the long-standing Swedish discussion of the head of en kopp kaffe @jnivre

Also, and I'm not sure that's relevant, at least in German this would not be swappable because a Weinglas would be a glass used for wine. But a Menge Formulare would be the same as a Formularmenge.

[shoowadoo]

But a Menge Formulare would be the same as a Formularmenge.

Do I understand you correctly? When these words are swapped, they form a compound with the same meaning? But for the case of Ein Glas Wein, when swapped, it no longer means a glass OF wine but rather a glass USED FOR wine.

Is this the rational for using appos in German here for Menge Formulare? Or for Stück Fleisch 'piece of meat', Haufen Schrott 'pile of junk', etc.

[dan-zeman]

Is this the rational for using appos in German here for Menge Formulare? Or for Stück Fleisch 'piece of meat', Haufen Schrott 'pile of junk', etc.

We don't know (and probably will never know) what the rationale was because the annotation was originally done by people outside UD, following guidelines that are probably not documented (at least I have no idea where to find them). But swappable or not, I think that ein Glas and Wein are not coreferential in the way we normally expect in apposition.

[jnivre]

Agreed. I don't know how relevant this is to the German case, but in Swedish the difference between a true partitive and a pseudo-partitive is that only the former uses a preposition to link the two nominals:

1. Ge mig ett glas av vinet = Give me a glass of the-wine (partitive)
2. Ge mig ett glas vin = Give me a glass wine (pseudo-partitive)

In 1, we treat "glas" (glass) as the head; in 2, we treat "vin" as the head; the relation is "nmod" in both cases. The choice to treat "vin" (wine) as the head in 2 is inherited from the old annotation of Swedish Talbanken and essentially takes the semantic head to be the syntactic head in the absence of any marker of subordination in either direction (no preposition). It does, however, introduce a cross-linguistic discrepancy with the annotation of English and other languages which use a preposition in both cases. (Compare, for example, French: un verre du vin vs. un verre de vin.) One can discuss whether this discrepancy is harmful or not, but one way to defend it is to say that Swedish (and maybe German?) has gone further than English (and French?) on the path of grammaticalization leading to a head flip.

If anyone is interested, @wcroft and I discuss this in a recent paper: https://aclanthology.org/2025.udw-1.1/

[LeonieWeissweiler]

Is this the rational for using appos in German here for Menge Formulare? Or for Stück Fleisch 'piece of meat', Haufen Schrott 'pile of junk', etc.

I don't know if this was the rationale, I was just mentioning it because you were asking if the elements can be swapped. But Stück Fleisch and Haufen Schrott are also swappable, so it sounds plausible that this was the reason for the annotation.

[shoowadoo]

We don't know (and probably will never know) what the rationale was because the annotation was originally done by people outside UD, following guidelines that are probably not documented (at least I have no idea where to find them).

I discovered this from the recent Bavarian treebank, and wanted to use consistent annotation with other related languages for Yiddish. I find the difference brought up by Leonie between ein Glass Wein and eine Menge Formulare interesting and a slightly convincing argument for appos for the latter. But then again, the fact that the two in reverse can form a compound with the same meaning is maybe a stronger argument for the use of nmod.

But swappable or not, I think that ein Glas and Wein are not coreferential in the way we normally expect in apposition.

I sense the other two nominals in examples like Stück Fleisch 'piece of meat' and Menge Formulare 'bunch of forms' also do not fulfill this condition, no? Or a shtikl broyt 'piece of bread', for a Yiddish example.