Function Dependents in UD v2
Crosslinguistic guidelines for the use of the functional labels such as aux, det, cop.
There is currently a lot of variation around this.
Representing lexical heads promotes crosslinguistic parallelism, but only if we can agree on what lexical heads are.
Chris’ thoughts on aux: This is necessarily going to have difficulties. Because of processes of grammaticalization, there are always going to be verbs that are “semi-auxiliaries” or heading in the direction of being “light” verbs or auxiliaries. At best, some languages have clear syntactic tests for what their auxiliaries are. For the goals of maximizing clarity, maximizing cross-linguistic similarity and making it more deterministic how to classify things, I believe that the best practical decision will be to define things like u-dep/aux and u-dep/cop narrowly. That is, classic auxiliaries like have, did are the only aux and we do not include the class to include all kinds of “light” verbs, such as the light/restructuring/complex predicate verbs of Romance languages with meanings like “begin”, “stop”, “go”, “come”, “try”, etc. For each language, we still need to come to some practical hard line of when to stop. This is Kim Gerdes’ “catastrophe” problem. E.g., for English, I’m tempted to say that the get passive should be called an auxiliary (u-dep/auxpass), but then it is not an auxiliary in the inchoative form where it means roughly “become”:
Copula
- Has been discussed in Uppsala 2015, see http://universaldependencies.org/2015-08-23-uppsala/copula.html
- The Uppsala solution has been criticized, and a change proposed in #329, see https://github.com/UniversalDependencies/docs/issues/329
Chris’ thoughts on cop: Range of predicates that are u-dep/cop: Similarly to aux, at one point in SD, we started allowing other traditional “copular verbs” to be treated as cop, such as become or appear. This seemed to end up very problematic for us and our users. So I would argue for treating all such words as content words and using cop only for the basic “to be” verb or verbs of a language. Analysis of oblique/adverbial predication with cop: I support what @nschneid and @amir-zeldes are proposing in https://github.com/UniversalDependencies/docs/issues/329 and their arguments, which considerably mirror the changes made between SD and UD v1. I think those were improvements. (I’ve also added a few more comments in that Github issue.)