Updates to Ancient Greek FEATS Options

[mr-martian]

So in the existing Ancient Greek treebanks, the following features are permitted:

Aspect=Imp|Perf, Case=Acc|Dat|Gen|Loc|Nom|Voc, Definite=Def, Degree=Abs|Cmp|Pos|Sup, Gender=Fem|Masc|Neut, Mood=Imp|Ind|Opt|Sub, Number=Dual|Plur|Sing, Person=1|2|3, Polarity=Neg, Poss=Yes, PronType=Dem|Int|Prs|Rcp|Rel, Tense=Fut|Past|Pqp|Pres, VerbForm=Fin|Gdv|Inf|Part, Voice=Act|Mid|Pass

In PTNK I currently have the following on definite articles and numbers:

PronType=Art NumType=Card NumType=Ord

Would the other treebank maintainers be OK with adding these features?

Are there any other features we should add?

Also, we should document how Tense and Aspect map onto the traditional paradigm names.

[dan-zeman]

@daghaug @gcelano

[Stormur]

Can I suggest adding InflClass? It should be pretty straightforward for Ancient Greek. By the way, comments and suggestions are welcome.

[mr-martian]

I also note that forms of τις don't have PronType currently, and the validator is objecting to my use of PronType=Ind.

[dan-zeman]

Apparently the other Ancient Greek treebanks do not use PronType=Ind for any word currently. It can be registered with the validator but ideally it should be then added to those words in the other two treebanks, too.

[daghaug]

I'd be happy to add PronType=Ind in PROIEL. It would be only for τις right, or am I forgetting something?

Dag

---

From: Dan Zeman @.***> Sent: 28 July 2023 09:57:48 To: UniversalDependencies/docs Cc: Dag Trygve Truslew Haug; Mention Subject: Re: [UniversalDependencies/docs] Updates to Ancient Greek FEATS Options (Issue #957)

Apparently the other Ancient Greek treebanks do not use PronType=Ind for any word currently. It can be registered with the validator but ideally it should be then added to those words in the other two treebanks, too.

— Reply to this email directly, view it on GitHubhttps://github.com/UniversalDependencies/docs/issues/957#issuecomment-1655225813, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ADCL7F5UJ3JOX5ZKPGOHUNLXSNWHZANCNFSM6AAAAAA2UCMWYI. You are receiving this because you were mentioned.Message ID: @.***>

[mr-martian]

Yes, in PTNK PronType=Ind is only on the 2 instances of τις.

[gcelano]

In the AGDT original files, articles have their own POS label ("l"), so this could be used for PronType=Art. Numerals have their own POS label as well ("m"), but there is no distinction between cardinals or ordinals. There are annotation inconsistencies, though, because numerals have sometimes been annotated as adjectives. There is also an issue with the distinction of articles vs pronouns in cases such as τιμῆς τῆς Πριάμου, where I take τῆς as a pronoun, but some annotated it as an article.

[mr-martian]

In PTNK, τῆς would always be tagged as DET. What are the conditions in the other treebanks for labeling it a pronoun?

[gcelano]

Well, an article is expected to agree with its governor in number, gender, and case (τιμῆς τῆς Πριάμου means honor, that of Priam)

[mr-martian]

I wonder how well that corresponds to the set of cases in PTNK where something tagged DET has a relation that is not det.

[mr-martian]

My code for exporting PTNK to the UD repo strips the features under discussion, but I don't think we've actually reached a consensus on what to do about them.

[dan-zeman]

OK, but then it needs a new milestone. v2.13 is over.