advcl:relcl for relative clauses modifying clauses

[nschneid]

• maybe they had bad teachers, which is a problem
• acl:relcl(had,problem)
• nsubj(problem,which)

Once again, we run into issues with deprels mixing POS/phrase type and grammatical relation. I think it's fair to say this is indeed a relative clause, but it's an adverbial rather than adnominal relative clause. So should it be advcl:relcl?

While some of the supplements arguably modify a nominal indirectly, and thus acl could be argued, this one is pretty clearly modifying a verb or VP/clause.

Originally posted by @nschneid in #762 (comment)

[nschneid]

The core group has decided to adopt advcl:relcl for such cases. Even though most relative clauses are adnominal, it would be technically wrong to use acl:relcl for non-adnominal uses of relative clauses. English guidelines will be updated accordingly. (advcl:relcl is already in use for some languages.)

[Stormur]

Nice. I have recently been stumbling very much on these kinds of "metaclauses" which recur often in some rhetoric styles.

Until now in Latin we have used advcl:pred, subsuming these modifying clauses under the phenomenon of secondary predication: in fact, there seems to be (at least in Latin) no other way to refer to a clause taken as a whole, instead than to a single argument (like in the case of a "depictive"), than another finite clause making use of a "generic" relative element in the neutral grammatical gender, which can be in singular or (more rarely) in plural, and is needed for reference. Though surely the notion of expansion that underlies both secondary predications and adnominal relative clauses creates grey zones. Actually, I think that this use of advcl can open up new ways to treat some cases often annotated as parataxis.

However, is the core group meant to be that of UD or only of English treebanks?

[nschneid]

We mostly discuss universal issues but have been doing a deep dive into English relative constructions. The hope is to then pursue typological guidelines for relative constructions.

[Stormur]

So will this be part of future UD amendments?

[nschneid]

Once we have better universal guidelines we can announce it as a clarification. Subtypes are technically optional and we're not really changing the meaning of acl:relcl as defined here.