Open rhdunn opened 11 months ago
Hmm, yes, it does appear the standard has changed. There is this issue:
https://github.com/UniversalDependencies/docs/issues/517
and the updated table is here:
https://universaldependencies.org/en/pos/PRON.html
Also, some of the features are out of date, I believe.
@rmunro any objections to me taking this on?
@rhdunn @nschneid @dan-zeman can you point me to where it was decided that the lemma of the pronoun mine should be the determiner my. I am definitely letting my Mordvin language knowledge influence me here, but it seems odd that the possessive my with a presumed lemma in I. This is my car. det(car, my)
Should serve simultaneously as the lemma of the pronoun mine. Whereas my would never be the lemma of itself. This car is mine. nsubj(mine, car)
Naturally, the mine word in alternation with my: mine eyes have seen... would point to I in the same way as my does.
That is to say that the referent of my and its archaic variant mine is the first person singular, the possessor, while the pronoun mine's referent is the possessum, car.
@amir-zeldes can you point me to the lemmatization of mine ="possessum" to my ="possessor" discussion?
It's becoming a party here in this tiny treebank!
People have gone back & forth several times about the PRON over the last couple years; the latest decisions are documented here, and they agree with @rhdunn 's analysis of these annotations:
@amir-zeldes can you point me to the lemmatization of mine ="possessum" to my ="possessor" discussion?
I think these issues are relevant:
I don't know about how it plays out in Mordvinic, but historically in IE languages, the possessives are a separate stem and not an inlfected form of the personal pronouns, at least for 1st+2nd person, so "my" was historically not be part of the paradigm of "I", and in some languages genitive forms of "I" coexist with these possessives (German mein-, meinet, Lat. "meus, meo" != "ego, mihi" etc.). For the third person the case could be made, but for simplicity we decided to keep separate lemmas for the whole PRP$ items (or substitutive possessive PRP).
Thank you @amir-zeldes for the pointers. At the moment, the languages I would like to cross-reference for this do not seem to have (m)any parallels to what might be analogical structures. So, I'll merely mention what kind of token types I was looking for: Hungarian: az enyém, a tied, a mienk... Moksha: mońńe, tońńe, mińńe... Etymologically these are associated with 1Sg, 2Sg, 1Pl in the two languages, but the references to these are az enyém/mońńe Case=Nom|Number=Sing|Number[psor]=Sing|Person[psor]=1 a tied/tońńe Case=Nom|Number=Sing|Number[psor]=Sing|Person[psor]=2 a mienk/mińńe Case=Nom|Number=Sing|Number[psor]=Plur|Person[psor]=1
Analogically, I would have seen English mine, yours, ours... as: mine Case=Acc,Nom|Number=Plur,Sing|Number[psor]=Sing|Person[psor]=1 yours Case=Acc,Nom|Number=Plur,Sing|Number[psor]=Plur,Sing|Person[psor]=2 ours Case=Acc,Nom|Number=Plur,Sing|Number[psor]=Plur|Person[psor]=1
And English my, your, our... as: I Case=Gen|Number=Sing you Case=Gen|Number=Plur,Sing I Case=Gen|Number=Plur
I understand, but I'm not sure it makes much sense for English, since "yours" being Acc,Nom
is not special to that pronoun - any noun in English is essentially Acc,Nom
, but we don't bother to note it, since only the personal pronouns proper distinguish case at all. Similarly, putting Number=Plur,Sing
on "mine" doesn't say much, and if they refer back to a coref chain that has both singular and plural members (which happens) it's not even clear that a human could exactly say what the grammatical number of the referring expression behind "mine" is. Changing the Number feature to a layered psor feature is conceivable, but again since there is no other grammatical number involved, I'm not sure it would be more helpful than confusing.
The independent possessive pronoun lemmas do not match the values in the Personal pronouns table:
Note: In the below output from my validator, the validation is matching lemma exceptions instead of the given general validation rule.
mine -> my
yours -> your
hers -> her
theirs -> their