Research update: Relative clause stacking
Work on relative clauses is coming along nicely. There’s a link on the antics part of the site to where I’m keeping all my notes on the matter, which should be updated every week or so around Monday. (Eventually, I’ll make a directory for all of my older stuff, as well.) The latest is a patch to my initial stab at embedding a domain restriction argument in the quantifier phrase. I’d flopped around the arguments so that the type of the relativized NP was quite silly looking and posed a problem to RC stacking. A sneak peek at where I’m at now (types are in parens because I’m too lazy to escape the angle brackets :-P):
[[every boy who left who came before John]] = λE(e,t)[λM(e,t)[ ∀x [(left’(x) & came’(x) & E(x) & boy’(x)) -> M(x)]]]
emma :: Feb.20.2007 :: misc, semantics, linguistics :: 3 Comments »
Of course, I can’t think of stacked relatives without thinking of Charles Fillmore…
If I’m reading this right, this is (let’s see how much HTML code this lets through! — corrections if necessary in a followup comment):
λE⟨e,t⟩ . λM⟨e,t⟩ . ∀x [(left′(x) & came′(x) & E(x) & boy′(x)) → M(x)]
But a function of type ⟨⟨e,t⟩,⟨⟨e,t⟩,t⟩ is what I might expect for [[every]], not [[every boy]] (with or without stacked relatives). Is “every boy who left who came before John” not a generalized quantifier? Or rather: what exactly is that E(x) doing in there?
Curses! That looked much better before the subscript HTML codes got eaten, and really, subscripting those semantic types was my main attempt to make it easier [for me] to read. Well, ignore the repetition of what you wrote, and skip straight to the question below it.
[…] I thought I’d just make a real post in response to Lance’s comment on my last post to clarify on the “fancy generalized quantifier” type (name courtesy of Polly :-P). […]