Dependency structure of "vad som är X" (sv)

[harisont]

I stumbled upon the sentence

(...) för att förstå vad som är nya trender i mode

and had to check what to do with "X vad som är Y".

Talbanken and LinES are both internally consistent in this respect, but they adopt two different approaches (query results here):

Talbanken

• ccomp(X,Y) or acl(X,Y) (depending on the POS of X)
• dislocated(Y, vad)

LinES

• obj(X, vad) or nmod(X, vad) (depending on the POS of X)
• acl:relcl(vad, Y)

I understand the rationale behind both approaches, but the guidelines make me think that the Talbanken solution is the way to go. Do you agree?

[jnivre]

This is a tricky one, and I think similar discussions have been discussed also for English. However, my position is that the complement of "förstå" is an interrogative subordinate clause, not a nominal object with a relative clause. This is less clear with a verb like "förstå", which can take a direct object as well, but I think it is the only reasonable analysis for similar examples with verbs like "fråga", "undra", which do not take nominal objects. Note also that it is not possible to replace "vad som" by "det som" in this case, without a change in meaning, which to me strongly suggests that the complement is an interrogative clause.

If we accept that it is a clause, there is still the question what to do with "vad som", where the "som" is this peculiar particle that is obligatory when the question word corresponds to the subject of the embedded clause:

jag förstår vad han har gjort
*jag förstår vad är på modet
jag förstår vad som är på modet

I have opted for an analysis, where "som" is treated as the grammatical subject of the clause (which is similar to how "som" is treated in relative clauses, where it is also obligatory if it corresponds to the subject) and where "vad" is treated as a dislocated doubling of the subject (structurally analogous to "Pelle" in "Pelle, han förstår inte vad som är på modet"). An alternative would be to treat "vad" as the subject, which would capture the parallelism with the object case, and to treat "som" as something else, possible "expl" as a kind of cllitic doubling.

[LarsAhrenberg]

I have been a bit influenced by the choices made in the English treebanks and I won't recapitulate the arguments for and against. As @jnivre points out there is an ongoing discussion starting with Italian here. I hope for some general recommendations before the next deadline.

[harisont]

Thank you both. Let's see where the discussions lead until the autumn then!