search

UniversalDependencies / docs

Universal Dependencies online documentation
http://universaldependencies.org/
Apache License 2.0
271 stars 245 forks source link

Consequential (conclusive?) conjunctions #471

Open msklvsk opened 7 years ago

msklvsk commented 7 years ago

There is a conjunction то (then) in Ukrainian, as in If something, then something. While then can be ADV, то is a pure conjunction (has no temporal meaning), but it's neither coordinating, nor subordinating: it connects the main clause to a preceding subordinate clause. How should we treat it? Our temp v2.1 solution is PART with special PartType connected by discourse. I believe, other languages have the same thing. Some analyze it as ADV, some as SCONJ, or even CCONJ + advmod as in Polish:

image

dan-zeman commented 7 years ago

In Czech the corresponding word is tak, it is tagged ADV and attached as advmod, see here: http://hdl.handle.net/11346/PMLTQ-IYAZ

I would incline to tag the Polish to ADV, too (in this context). If it stays CCONJ, its relation should be cc, although there is no coordination.

dan-zeman commented 7 years ago

(Another possible consequential adverb in Czech is potom. Same analysis as tak.)

dan-zeman commented 7 years ago

Edit: It turns out that some trees returned by the query above contain tak tagged CCONJ. So we have a consistency issue that should be fixed.

msklvsk commented 7 years ago

But то is not an adverb. You cannot ask “when? — то.” There are adverbs that can be used in similar context, but то is just a linker.

dan-zeman commented 7 years ago

I see your point but I don't know what would be the best (or least evil) solution here. In Czech, tak is an adverb, but it is originally a demonstrative manner adverb. You would ask jak (how), not proč (why). So in this context it is more of a linker, too. It is a corelative element that refers to (or represents) the whole adverbial clause that specifies the condition. The fact that it is a pro-form representing an adverbial seems to me to support calling it pronominal adverb (regardless the fact that in other contexts in Czech it is a manner adverb). Could we perhaps use the same reasoning for Ukrainian and Polish то/to?

jnivre commented 7 years ago

I definitely think adverb is the best solution here. There are many discourse particles that cannot be used to answer when/why/where-type questions but are nevertheless classified as adverbs. It also seems like the best solution for cross-linguistic consistency.

dan-zeman commented 6 years ago

Leaving this open because of the unresolved inconsistency in the Czech data. Setting a new milestone.