RST

[Maite Taboada]

At 21:23 21/01/99 +0200, Gisela Redeker wrote:
>Dear Maite,
>
>I'd be very interested in seeing how you have applied RST to
>conversation data. The few attempts that I've seen in the past all
>involved using RST for turn-internal structures and supplementing
>other kinds of relations for the exchange structure.
>
>Please let us know when your paper becomes available on your web
>page!

The paper is now available, a version in postcript format, and a slightly
shorter one in RTF. Go to the link that says "Publications" (address of the
web page in the signature of this message). The paper is titled "Rhetorical
Relations in Dialogue: A Constrastive Study".

Now, responding to Gisela: yes, indeed, people have used RST mostly for
turn-internal structures, and that's where I started. The paper includes an
extensive analysis of those. But then I realized that what we (I) do when
we analyze a conversation is partly to look at it as a product, as a text.
The speakers are building through their interaction a text that I then take
as a finished product. Those conversations showed a global coherence that I
wanted to explore using RST. I then saw how you could use relations of
evaluation, solutionhood, restatement, etc., to describe the relationships
among spans of the dialogue that did not necessarily begin and end in each
turn, that is, relations whose nuclei and satellites included text/talk
that spanned across turns.

Except for a few cases where the connectedness constraint is broken,
because there are instances of self-talk, or for other reasons, I did
manage to analyze all 60 of my dialogues, in both languages, English and
Spanish, as whole texts. There are differences in the types of relations
and the number of relations that are present in each type of analysis
(number one being the analysis of turns in isolation, and number two being
the analysis of whole conversations). That's why I've began to wonder
lately whether we could classify RST relations in relations that tend to be
local (condition, concession, e.g.), and relations that tend to be global
(evaluation, solutionhood). But it could also be the case that in my corpus
those are the global/local relations, and you could find different
distributions in other genres/types of conversation/modes.

Cheers,

- Maite

		Maite Taboada
Department of Linguistics	Ph:  (403) 492-3480
4-20 Assiniboia Hall		Fax: (403) 492-0806
University of Alberta		e-mail: mtaboada at ualberta.ca
Edmonton, Alberta		or: 	flingz7 at sis.ucm.es
T6G 2E7 Canada			http://www.cs.cmu.edu/~taboada

Note: The area code will change to (780) Jan. 25, 1999