conclusion relation

[Eduard Hovy]

Hi Gisela, and all,

At 6:03 PM +0200 8/21/08, Gisela Redeker wrote:
>Where I differ from Ed, then, is the consequence drawn from the fact that
>text segments often or always have (more or less obvious) multiple relations.
>He argues that RST should not conflate these components. While I agree that
>they should be recognized and modelled as separate components, I do feel
>that the conflation in describing the functional structure of a text is a
>strength of RST, not a flaw. It allows the analyst to describe the interplay
>of levels in the achievement of the writer's purposes. In my experience,
>focusing on the writer's
>purposes almost always brings out one of the concurrent relations between
>segments as the most salient one given those purposes. Of course there are
>cases of parallel analyses, but they are as often ambiguities within a
>metafunction than across.

You raise an interesting point.  I just want to see if I understand 
it accurately?  (I'm afraid I don't understand the use of the word 
"conflate" here.)

Accepted:
1. Adjacent text segments can be (and often are) related in multiple ways
    simultaneously (typically, one from each metafunction)
2. These individual analyses should be recognized and represented separately,
    to keep things clear

In order to achieve point 2, one needs a labels for each relation, 
and the labels must be unique to their relation, no?  That is, one 
needs a name like "Relation X" or perhaps "Relation X (ideation)" 
specific to each relation?  (Otherwise, one can't represent the 
multiple relations uniquely.)

The collection of all possible relations (probably sorted into the 
three metafunctional groups) would then constitute the universe of 
discourse.  An 'exhaustive' theory would create one label for each 
relation for each metafunction.  Some other theory might merge two or 
more metafunctions into one label.  RST is not 'exhaustive', in two 
ways:
- sometimes an RST relation is 'hybrid' in that it carries two metafunctions,
- sometimes there is no RST relation for the relation.

Is it correct that the difference in our opinions is the following?
1. I believe RST would have been clearer had it been 'exhaustive': 
had it distinguished all relations accordingly, namely one unique 
relation label for each possible metafunctionally unique relation 
between segments.
2. You believe RST as it is now is better.

Now my question: When one works with a set of labels that are 
hybrids, then how can one describe the situation when only one 
metafunction alone obtains, if there is no 'pure' label for the 
relation found there?  I.e., all relations one would tend to place 
there also bring in connotations from the other metafunction(s)?  (In 
other words, how does one in practice bring out one of the concurrent 
relations?)

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.

I hope I have described things accurately.  Thanks for the 
opportunity to try to do so!
Regards,
E


-- 
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