extended relation set
bill_mann at SIL.ORG
Mon Oct 29 19:04:51 UTC 2001
Thanks for your question about RST relations.
The extended MT relation set of Mick's tool is from me, and it is not
published. It should be identical to the one that is described in the
[tools] section of the RST website.
When we were putting RST together, we held to a policy of not defining a
relation just because "surely, surely we will someday find a text that has
this one." It is a somewhat conservative, empirically driven policy. In
the additions that define the extended set, there are some for which we have
seen only a few examples. In one case, the unconditional, I have seen only
one example, but it is a very clear example.
The hope in defining RST was that people would define relation sets to fit
their research needs. That has not happened much yet. Clearly in the area
of logic or argumentation there are possibilities for much more delicacy.
There are other places that would benefit from refinement as well.
My only restriction would be that relations that have names must also have
definitions. The most meaningful names are mostly polysemous, and often we
do not notice that at first. For extreme cases, certain words should never
be used as relation names because they suggest so many different ideas to
----- Original Message -----
From: "Renate Henschel" <rhenschel at UNI-BREMEN.DE>
To: <RSTLIST at LISTSERV.LINGUISTLIST.ORG>
Sent: Friday, October 26, 2001 2:31 AM
Subject: [RST-LIST] extended relation set
> In Mick O'Donnell's RST tool, there are two different sets of
> possible RST relations differentiated:
> - the classic MT (Mann and Thompson) relation set
> - the extended MT relation set
> I suppose that the classic set is identical to the set of relations
> defined in Mann and Thompson 88. Now I am looking for a citeable
> source for the "extended relation set". I hoped to find that set
> at the RST homepage, but I could'nt find it there.
> Which paper does define the extended relation set? If this paper
> is not available on the net, I would appreciate if the extended
> set could be made public at the RST home page.
> Renate Henschel
More information about the Rstlist