Reply on the problem with an RST analysis

[William Mann]

All:

I would agree with the comment that the text is incoherent.  I don't understand
it as given, and I suspect that the problem is not with the translation to
English.

Yes, more context is needed, and perhaps more background knowledge of the
situation and topic as well.  With more context, it might turn out that some of
the items link up to the left and another group to the right, with no connection
in this interval.

Also, some of the units chosen for analysis are restrictive relative clauses or
prepositional phrases.  It is true that RST can analyze some relationships
within sentences, but I don't know that RST (at least what Nick Nicholas calls
CRST, classical RST, per the 1988 paper in Text 8(3) ) is reliable at those
levels.

On this text, metaphor is also adding to the difficulties.  It might be more
instructive to abandon this text and try to investigate the same issue with some
other small whole text.

I wish you the best of success in your research.

Bill Mann



______________________________ Reply Separator _________________________________
Subject: DISCOURS Digest - 19 Oct 1999 to 20 Oct 1999
Author:  <LISTSERV at LINGUIST.LDC.UPENN.EDU> at Internet
Date:    10/20/99 11:00 PM



..


----------------------------------------------------------------------

Date:    Wed, 20 Oct 1999 10:08:32 +0100
From:    Luuk Lagerwerf <l.lagerwerf at WMW.UTWENTE.NL>
Subject: Re: DISCOURS Digest - 15 Oct 1999 to 19 Oct 1999

Hi Holger,

In reply to your RST problem, I would say that the text is incoherent.
This is not an option in RST analyses, but the topic that is treated as
known to the reader in (2): "the result of six months of negotiations",
seems to be introduced afterwards in  (5) "When finally ... is
announced".  But maybe I'm missing context before (1), or some stylistic
trope has been used that I don't recognize in the English translation.

Kind regards,

Luuk Lagerwerf


>
> (1) What else but shame should one feel
> (2) in face of the result of six months of negotiations,
> (3) which started with a nice gesture
> (4) only to end soon in woeful horse trading.
>
> (5) When finally in the next few days at the negotiations in Washington
>     the sum of 6,9 billions DM is announced, =
>
> (6) this means: it's sink or swim.
> (7) Because, for the former forced labourers every week counts.
>
>
> (6) is "Vogel, friss oder stirb" in German, which means "bird, eat it
> or die".
>
>
> Okay, now here's the problem: decide to which unit (5-6) relates, and
> with which relation. I tend to say (1-4) and "Joint" (or in D. Marcus
> terms "Topic Shift"), although one could argue that in (5) the
> "result" from (2) is ELABORAT-ed ("finally"). This however would
> imply that (5-7) are sub-ordinate to "feeling shame" (1) and I am
> not sure about this reading (i.e. I am not sure whether the author
> wanted to convey it, my own feelings are not an issue here).
>
> Apart from this, I would say, 1 is related to 2-4 via a CIRCUMSTANCE
> relation, so the conditional in (5-6) would also be a part of this
> CIRCUMSTANCE which does not look too convining to me.
>
> Any comments anybody ?
>
>  Holger
> -- =
>
> ---          http://www.coling.uni-freiburg.de/~schauer/            ---
> "Kann mir jemand kurz das Anliegen der newsgroup erkl=E4ren?"
> "Wahrheitssuche, eingeschr=E4nkt auf das Thema Fahrrad."
>                   -- C. Karl und Bernd Sluka in de.rec.fahrrad
>
Luuk Lagerwerf
Language and Communication
University of Twente
+31 53 489 35 67 (office)
+31 34 351 38 81 (home)

------------------------------

Date:    Wed, 20 Oct 1999 20:32:43 +0200
From:    Gisela Redeker <G.Redeker at LET.RUG.NL>
Subject: Re: A problem with an analysis in terms of RST

I agree with Luuk Lagerwerf that more text is needed to
adequately answer Holger Schauer's question about the relation
between 1-4 and 5-6 below (though I hesitate to agree with Luuk
in calling this text 'incoherent'!):

> (1) What else but shame should one feel
> (2) in face of the result of six months of negotiations,
> (3) which started with a nice gesture
> (4) only to end soon in woeful horse trading.

> (5) When finally in the next few days at the negotiations in Washington
>     the sum of 6,9 billions DM is announced, =
> (6) this means: it's sink or swim.
> (7) Because, for the former forced labourers every week counts.

Considering the genre (news article), I'd guess 1-4 is the
evaluative (or commentary) end of the last (or only) "take" of the
article, followed by what Van Dijk might call an "expectation"
segment (probably the final paragraph of the article).

If that is true, we should not expect a strong rhetorical relation
between these two paragraphs. As I have argued elsewhere
(Redeker 1996/forthcoming), RST lacks special 'machinery' to handle
genre-specific structures (where sequential relations between text
segments are defined in terms of the structural elements expected,
and not in terms of content or argumentation), and thus has to go
with the nearest semantic or pragmatic relation-- in this case I'd
say "SEQUENCE"--with the understanding that I assume 1-4 is the
EVALUATION satellite within a larger segment,
to which 5-7 can then form the 'next' in the sequence.

This analysis would seem particularly appropriate if the earlier
global strategy was also sequential (chronological report). If this is
not the case, other possibilities might offer themselves, such as
INTERPRETATION (with 5-7 as the satellite introducing the
wider/new perspective) or a simple JOINT if the article is built up
like a series of (not chronologically ordered)
observations+commentary. The lattter is the option Holger
discusses in his original mail:

> Okay, now here's the problem: decide to which unit (5-6) relates, and
> with which relation. I tend to say (1-4) and "Joint" (or in D. Marcus
> terms "Topic Shift"), although one could argue that in (5) the
> "result" from (2) is ELABORAT-ed ("finally"). This however would
> imply that (5-7) are sub-ordinate to "feeling shame" (1) and I am
> not sure about this reading (i.e. I am not sure whether the author
> wanted to convey it, my own feelings are not an issue here).

I'm not sure 2-4 should be subordinate to 1 (see below). But the
attempt to link individual units across others prompts a general
comment on (my understanding of) RST-methodology:
What I see as very important and very useful in RST is the focus
on local connectedness: For any unit, look first how it is connected
to its immediate neighbors and how that cluster may be related to
neighboring clusters (in the process of analysis, you may need to
go up and down a few times between these, as the more global
structure can affect the relative centrality (esp. nuclearity) of a unit
inside a cluster). This principle is a useful guideline for the analyst,
but more importantly it makes a lot of sense when we think about
actual (online) text processing (which RST is not trying to model;
but it's nice that it is compatible with it).

> Apart from this, I would say, 1 is related to 2-4 via a CIRCUMSTANCE
> relation, so the conditional in (5-6) would also be a part of this
> CIRCUMSTANCE which does not look too convining to me.

This is VERY difficult to analyze without further context. It sounds
a bit like the article has just reported on someone being upset and
ashamed about the result. In any case, 1-4 could not be the first
units of the text, and 1-4 as a whole might be a RESTATEMENT or,
if the details in 2-4 have not been stated before, an
INTERPRETATION of the preceding cluster. Given some such global
analysis, I'd analyze 1 as an EVALUATION satellite at 2-4, which in
turn consists of a CONTRAST relation (3-4) that forms an
ELABORATION satelite to 2.

As for 5-7, I'd analyze 5 as a CIRCUMSTANCE satelite left-attached
to 6, with 7 right-attached to the combination 5-6 as an
INTERPRETATION satelite (introducing a new perspective).

Again, this is very speculative because of the missing co-text
(apart from being a first-off analysis that would need a second
look after one or two days and/or discussion with co-analysts). But
maybe Holger or some other list readers will find these useful to
mull over and criticize...

Happy analyzing!
Gisela


Gisela Redeker, Professor, Dept. of Language and Communication
University of Groningen,  P.O.Box 716,  NL-9700 AS Groningen
tel:  +31-50-3635973/-5858     fax:  +31-50-3636855
e-mail: G.Redeker at let.rug.nl   http://www.let.rug.nl/~redeker/

------------------------------

End of DISCOURS Digest - 19 Oct 1999 to 20 Oct 1999
***************************************************