I was wondering what Person=Auto is supposed to mean? It occurs 31times with Breton verbs but it is not documented and it does not appear in any other Celtic treebank (or, in fact, in any other treebank).
This is the autonomous/impersonal form.
âAnother shared trait in the verbs is the presence in the paradigm of the âimpersonalâ or
âautonomousâ verb form. Basically, all Celtic languages possess an impersonal form for
each tense which is neutral as to the person and number features of the subject. So Welsh
dysg + ais âI taughtâ in fi rst- person singular contrasts with the impersonal form dysg +
wyd âone taughtâ. While this form can often be translated as a passive (âis taughtâ), the ending also occurs with intransitive verbs, as with Irish tĂĄthar âthey/people areâ. The impersonal paradigm is an important inherited feature from Indo- European, since it partakes of the *- r ending which also appears in Italic, Tocharian and Hittite. The actual usage of these forms has diverged significantly over time (in Welsh these have become rather literary constructions, but they are everyday forms in Irish), but the presence of a special verbal inflection for an unspecified subject is another particular feature of Celtic.â (Fife, 1993)
It appears in:
Voice=AutoPerson=0Person=0 in Scottish Gaelic.Calling all Celtic UD contributors: @tlynn747 @colinbatchelor @jheinecke @kscanne :)
I think that there are some similar constructions in Finnish too (see e.g. description of passive and zeroth person).
Polish also has Person=0 for a similar purpose, e.g.:
Question for my understanding: if such impersonal forms coincide with personal endings, as it happens in Latin (with the 3rd person, often passive), shouldn't one just use the same morphological label as for that form? This somehow reminds me of the recent discussion about deponent verbs (#713).
It's Person=0 in Scottish Gaelic also. This construction doesn't exist in the Manx corpus; usually some kind of circumlocution is used to convey this meaning.
Question for my understanding: if such impersonal forms coincide with personal endings, as it happens in Latin (with the 3rd person, often passive), shouldn't one just use the same morphological label as for that form? This somehow reminds me of the recent discussion about deponent verbs (#713).
In Irish at least, the impersonal forms usually do not coincide with any personal form.
Question for my understanding: if such impersonal forms coincide with personal endings, as it happens in Latin (with the 3rd person, often passive), shouldn't one just use the same morphological label as for that form? This somehow reminds me of the recent discussion about deponent verbs (#713).
In Coptic we have impersonal 'passives' that look exactly like 3rd person plural active, so as you suggest, we tag them like 3rd person plural, even though they can have a by-agent phrase:
If there's no agent, you can't distinguish "they killed him" from "he was killed".
But the Polish forms in the link above are different - they use a special morphological form dedicated to having an impersonal agent (in fact you cannot specify the agent argument with that form). I think if the form is non-identical with other morphological forms, it deserves its own person value. Ancient Egyptian also has such forms, though they did not survive into Coptic, and Hausa is the same:
+1 for standardizing the value for this. Anything is fine IMO, I suppose Person=0 has the advantage of keeping all values numeric?
Similarly for Welsh. It is different from the 3rd Singular and a verb in the impersonal form does not trigger soft mutation on a indefinite direct object. And there is no pronoun, so these .. From a semantic point of view it is rather close to French "on + Verb" or German "man + Verb", with the difference, that in French or German in this case the verb is in 3rd Singular
I've updated the Breton to be Person=0 (from Person=Auto) in https://github.com/UniversalDependencies/UD_Breton-KEB/commit/dd7ec81e3b962e3ab78977e95feaf5384ed9e34f. But before closing this issue would like to get feedback regarding the Irish treebank.
Thanks! It sounds to me like Person=0 is the right solution here (we added it in v2 UD guidelines, inspired by Unimorph, although their specification did not mention Polish, Ukraininan or Celtic). Let's see what @tlynn747 thinks about it.
I have checked Person=0 as valid with VERB and AUX for Breton here, so the validator no longer reports the feature as an error. (But it still reports some langauge-specific relation subtypes.)
I've updated the Breton to be
Person=0(fromPerson=Auto) in UniversalDependencies/UD_Breton-KEB@dd7ec81. But before closing this issue would like to get feedback regarding the Irish treebank.
It is interesting to note that the Finnish and Estonian treebanks use the feature Voice=Pass when dealing with what the traditional grammars call passive, but whose meaning would, indeed, be a near if not perfect match for the impersonal Person=0 addressed here.
The Finnish and Estonian forms are distinct from the 1â3 personal forms. The structure does not work with an agent, except in the spoken language, where it is regularly used to indicate the first person plural in Finnish.
@rueter : But 3 Finnish treebanks (out of 4) do use Person=0. Estonian does not use it, though.