RST structure evaluation

[Arthit Suriyawongkul]

Dear RST list,

I'm an MSc student currently doing master project on automatic
summarisation,
and it has to deal with RST discourse structure.

Say if we have a set of RST relations, and a parser (machine or human)
which get plain text as its input and has an RST discourse structure as
its output.
Does there any widely accepted way to evaluate a correctness of RST?
(I realize that there could be many RST interpretations for one same text.)

Does anyone has a pointer to resource on this please?

Note: I'm not going to write my own RST parser. Instead, I will use the
structure
already avaible in RST Discourse Treebank as an input of the summariser.
Anyway, if possible, I like to see how reliable of the input it is.

Thank you :)

regards,
Art