conclusion relation

[Eduard Hovy]

Hi John, and all,

At 10:02 AM +0200 8/23/08, John Bateman wrote:
>>It seems to me the analyst has more freedom to describe the interplay
>>of levels (= metafunctions) when he/she can describe each relation
>>separately, using an 'exhaustive' set.
>
>The drawback is that many of the constraints that come from
>RST (nuclearity, adjacency) begin to be weakened and one is
>more or back with something that looks like a conjunctive
>relation analysis, which is significantly underconstrained
>structurally.

Just to confirm: Assuming one has an 'exhaustive' list of all 
possible relations for each metafunction (Ideational, Interpersonal, 
and Textual), then Conjunctive Relations allow one to combine one of 
each set in every instance of analysis, yes?  (Even though sometimes 
it might be very hard to decide which particular relation obtains for 
one or two of the metafunctions.)

If my understanding is right, then your interesting comment above 
suggests that it is not the case that one can find in natural text 
ALL possible permutations of one of each metafunction; that is, that 
there are some specific Id-Int-Txt triples that simply do not occur 
in text, ever.  (And RST of course would reflect this, by not 
providing relations for such impossible combinations.)

While this seems to me plausible as an argument, has anyone ever 
actually gone to the (not inconsiderable) trouble to test the claim 
empirically?  It's hard to prove a negative, of course, but if one 
were to annotate a lot of text and were to find many examples of 
Ideational relation X and Interpersonal relation Y but absolutely no 
co-occurrence of the two, then perhaps one could conclude that the 
X-Y combination was not feasible for some reason.

And then, of course, the question is: why not?  Surely the 
Interpersonal is sufficiently 'independent' of the Ideational not to 
be influenced/influenceable it it?  (It may be that constraints of 
nuclearity or adjacency rule out some combinations, certainly: that 
is, while there would in principle be an X-Y combination, the fact 
that X is Nucleus-first and Y is Satellite-first makes their 
combination impossible, for example.  If so, it should be relatively 
easy to produce impossible X-Y pairings synthetically, 'in the lab', 
so to speak, and then hunt for them in text.)

Absent such examples, I want to remain agnostic on the question of 
whether there are some triple combinations that never occur in 
nature...I'd be dead eager to see a proven answer -- one way or the 
other.

-- 
Eduard Hovy
email: hovy at isi.edu            USC Information Sciences Institute
tel: 310-448-8731            4676 Admiralty Way
fax: 310-823-6714            Marina del Rey, CA 90292-6695
http://www.isi.edu/natural-language/nlp-at-isi.html