Coherence?

Holger Schauer Holger.Schauer at GMX.DE
Fri Dec 17 11:40:06 UTC 1999


>>>>"VI" == Viatscheslav Iatsko schrieb am Fri, 17 Dec 1999 17:53:28 +0700:

 VI> According to it coherence is based upon logical
 VI> relations between sentences, such as synchronic, diachronic, and
 VI> causative-consecutive. These logical relations constitute the
 VI> background of three types of discourse: description, narration,
 VI> and reasoning. Cohesion and coherence relate to each other as
 VI> surface and deep levels of discourse structure.

I whole-heartly agree to this view of coherence. Cohesive devices,
such as the ones mentioned, give us one notion of "relatedness". But
it does not give us any notion [1] of the structure of a text. What I
have in mind is the typical example of a scrambled text (i.e., one
takes a coherent text and then scrambles the order of its sentences):
one can still find the correct coreferences (although this might not
be true for pronoun resolution) but the text is no longer
coherent. There are other *structural* constraints which are not
addressed by cohesion (or at least not by the mentioned cohesive
devices). If certain constraints are met, one can say: hey yes, this
sentence is related to that one by this relation etc. I believe this
is what coherence is about. Isn't it ?

I believe that cohesion is a necessary but not sufficient condition
for coherence. I.e., you might have cohesion without coherence. I'm
not sure about the other way 'round (that's the reason for the query
for references in my reply to Bill Mann).

Holger

BTW: Should we take this discussion to DISCOURS ?

Footnotes:
[1]  I'm overgeneralizing here. Work on coreference, for example,
shows that in normal (i.e. coherent :-) text one can come up with
a structure solely based on coreference. E.g.,

@InProceedings{hahn-strube:acl97:centering-large,
  author = 	 "Hahn, Udo and Strube, Michael",
  title = 	 "Centering in-the-large: Computing referential
		  discourse segments",
  booktitle =	 "Proceedings of the ACL-97",
  year =	 1997,
  pages =        "104-111",
  shorttitle =	 acl-97s,
  keywords =	 "text cohesion, anaphora, anaphora resolution,
		  centering model, text coherence, discourse structure",
}

--
Holger Schauer                         CLIF - Computational Linguistic Lab
                                       Freiburg University, Germany



More information about the Rstlist mailing list