[developers] Subordinating Pairs Analysis

Ann Copestake aac10 at cl.cam.ac.uk
Sat Jun 17 14:25:22 CEST 2017


Hi Kristen,

I can discuss the way the MRS might look, though not the details of how 
you get there.

If there's a semantic relationship between the two clauses, then there 
needs to be some sort of two-place predicate taking the LTOP of each 
clause as an argument (usually via a qeq).   If the two elements of the 
pair always go together, and there is a restricted range of options, 
this two-place predicate might be the only element of the semantics.  If 
both elements are adverbial, the semantics might have to be associated 
with the construction rather than trying to do it via unusual semantics 
for an adverb.

Looking at the ERG demo and delphin-viz, it seems that if_x_then is used 
for a range of situations, including ones without any lexical marking - 
e.g.,

"Had I slept, it rained." (actually I find that ungrammatical, but never 
mind ...  "Had I slept, it would have rained." is fine)

In terms of the actual semantics, one could say there are two things 
going on with if_x_then - one is a causality relationship and the other 
is a hypotheticality marking.

"I slept, so it rained."

is just causality.   So one could analyse

if X then Y.

as (schematically)

cause(hyp(X),hyp(Y))

and

X so Y

as

cause(X,Y)

I don't think this would be a good idea for English (too much 
decomposition, so it probably doesn't capture the nuances), but it might 
be more convenient for other languages.

It is not the case that we can always capture the meaning directly for 
English.  For instance:

"I slept and, as a consequence, it rained."

implies causality, but we won't capture that directly in the MRS.  I'd 
say that what's going on is that `and' gives a two place relationship of 
the right form, but highly underspecified.  "as a consequence" means it 
has to be interpreted causally.

In context:

"I slept and it rained."

can do the same thing.

To sum up, what I'm saying is that I think you'll always want some type 
of two-place clausal connective, but it might be underspecified to some 
extent with additional meaning conveyed via additional predications on 
individual clauses.

All best,

Ann





More information about the developers mailing list