Days of the week / months

[AngledLuffa]

Are days of the week or months subject to a universal standard for PROPN vs NOUN, or is it on a per-language basis?

If it's on a per-language basis, should we at least come up with standards for each language as opposed to having different values in different treebanks?

For example, in English they all seem to be PROPN, and it is consistent across treebanks.

In German, GSD tags them all as PROPN, HDT tags them all as NOUN

in Spanish, GSD again tags them all as PROPN, AnCora tags them all as NOUN

I don't mind seeing them switch to NOUN if that's better in some languages, but it'd be better for the treebanks in Spanish, German, or other languages with conflicts to all agree

[nschneid]

My guess is that it should vary across languages, but yes, ideally should be standardized within languages.

[AngledLuffa]

That sounds fair to me... but how to agree on treatments in those two languages in particular, and how to make those updates?

(We already started the German discussion here: UniversalDependencies/UD_German-HDT#14)

[nschneid]

Well it's a question of whether they pattern more like nouns or like proper nouns in the language.

For example, in English proper nouns (including weekdays and months) generally do not take articles, whereas other nouns do. But in French the days of the week do take articles, so that could be a reason to prefer NOUN.

[jnivre]

It is true that weekday names overwhelmingly do not take articles in English (as shown by @amir-zeldes in a different thread). But it is also true that they inflect for number, which proper nouns normally do not:

Tell me why I don't like Mondays.
The council normally meet on Thursdays.

On the more general point, I agree that this should be standardized on a per language basis.

[AngledLuffa]

You can inflect for number in English for other proper nouns as well. "The Bauers are arranging a ___ event next week" Hockey example: "The Tkachuks are elite at getting under their opponents' skin".

[dan-zeman]

I think in languages like Czech or Spanish a good approximation will be to simply see whether the orthography of the language requires such a word to be capitalized (by itself, not because it is part of a multiword named entity or starts a sentence). Especially if it is difficult to find criteria that will not depend on written form, such as articles (not applicable in Czech).

Unfortunately, this approach does not help in German, where all nouns are capitalized.

[AngledLuffa]

Why is domingo hard? Do you want me to go through looking for person vs day domingos?

[dan-zeman]

Why is domingo hard? Do you want me to go through looking for person vs day domingos?

If you have capacity for it, it would certainly help. I originally had domingo in the regex and saw in the diff mostly instances like Santo Domingo or Plácido Domingo. Looking at capitalization would probably help, but I initially ignored it to circumvent typos and stuff.