Closed nschneid closed 5 years ago
Actually, I guess cringing is an act which requires a cringer, and inducing is causation, so something like
[a_E [(IMP)_A cringe_P inducing_D (presentation)_A]_E presentation_C]_P [by_R the_E mayor_C]_A
or
[a_E presentation_C]_P [(IMP)_A cringe_P inducing_D (presentation)_A]_D [by_R the_E mayor_C]_A
?
Considered and rejected: [a_E [cringe_P inducing_D]_E presentation_C]_P
The solution, though somewhat unintuitive: [a_E]_H- [cringe_P inducing_D]_H [presentationC]-H This avoids a tricky issue of defining which modifier scenes are D and which are H.
TBD: the argument structure of "cringe-inducing".
Should we include implicit/remote participants of nominalized events/states like "cringe-inducing" or "express discontent"?
We need to ask the FrameNet people about the typology of DNIs vs. CNIs, and can we define a well-defined sub-class of implicits to annotate, deferring the annotation of others.
I talked with Miriam. It seems DNIs are just as tricky to get as INIs. We'll have to do strictly with CNIs for now.
p. 18 addresses noun scenes. One of them has an adjective ("accurate") tagged as D. What about an adjectival modifier that is more scene-like, e.g. "a cringe - inducing presentation by the mayor"? Would this be
[a_E [cringe_A inducing_P (presentation)_A]_E presentation_C]_P [by_R the_E mayor_C]_A
with a remote to indicate that the head of the adjectival modifier is essentially an argument of it?
Another possibility that avoids using E for a scene modifier:
[a_E presentation_C]_P [cringe_A inducing_P (presentation)_A]_D [by_R the_E mayor_C]_A