coherence relations between large spans of text

[Anna Kazantseva]

Dear list members,

I  a Ph.D. student at the University of Ottawa, working on coherence of
short stories (classical fiction) from the perspective of Computational
Linguistics. I wonder if any of you are aware of any papers/books/other
pointers about what sort of coherence relations can hold between relatively
large segments of text, either in RST or in another tradition.

I know that RST does not make a distiction between large and small spans  of
text.  But  it seems to be the case (looking  at Discourse TreeBank) that
people agree less and less on specific coherence relations and even on the
shape of the tree the higher up the tree they work and the larger text spans
get.  With fiction this is a real problem because texts are relatively
large. I am looking for alternative ways to represent coherence between
large segments of text (paragraphs and  larger),  but so far with little luck. 
Thank you in advance for any pointer. I will post a summary of the responses
I get.
Regards,

Anna Kazantseva