search

UniversalDependencies / docs

Universal Dependencies online documentation
http://universaldependencies.org/
Apache License 2.0
273 stars 248 forks source link

Voice: passives and causatives #998

Closed LanguageStructure closed 12 months ago

LanguageStructure commented 12 months ago

Some languages have a passive marker that cooccur with a causative marker: Bororo: aogwa-do-dy hide-CAU-PASS 'caused to hide' (be interrupted).

Since causatives and passives are rarely in complementary distribution, I wonder why they are both voice markers UD. Is it possible to use both without creating a new feature?

amir-zeldes commented 12 months ago

This issue may be helpful: https://github.com/unipv-larl/UD4HL/issues/9

dan-zeman commented 12 months ago

The de-facto standard solution to this in UD is to define not a new feature, but a language-specific value of the existing feature Voice=CauPass; see e.g. Turkish.

LanguageStructure commented 12 months ago

Thank you! I shall stick to Voice=CauPass.

Stormur commented 9 months ago

I had to think back to this issue while reading a passage in Dixon's Basic Linguistic Theory 1 about syntactic terminology (p. 235):

Voice is the traditional term for derivations which reduce valency—passive and now also antipassive. It is preferable to restrict ‘voice’ to this sense, and not extend it to valency-increasing derivations such as causative and applicative, or to reflexive and reciprocal when they change valency. (Some linguists do accord a wider meaning to ‘voice’ and do so in a variety of different ways, which may lead to confusion.)

So, this could in fact lead to acknowledge different features for values like Causative and Passive, based on whether valency is positively or negatively affected, and from the data like for Bororo we do get a confirmation of this (but it is already well known that this happens periphrastically for many languages, e.g. Italian mi sono fatto vedere 'I made see myself', using causative fare 'to make/do' and "reflexive" strategy).

Possible tags could depend on other terms like (de)transitivizing (also used in Dixon's work), maybe also covering possibly morphologically expressed ambitransitivizing strategies (one example, very dear to Dixon, could be Fijian), and others.

In my opinion this should be seriously considered, and: might a similar feature split be not so drastic to be still part of UD 2?