Relations that are seldom or never signale

Bill Mann bill_mann at SIL.ORG
Tue Jan 18 20:41:52 UTC 2000


Amazing.  Twice in one day.

I want to reply to Alistair Knott's message about multiple levels.

I have read in a number of published papers that RST does not permit multiple
analyses, multiple relations holding across the same pair of spans.  (I have
also read that RST analyses are all trees [NOT SO] and that it does not permit
relations to cross over some span that is not included in the immediate spans
being related. [NOT SO AGAIN.])

In RST as I understand it, the analyst can affirm more than just a simple
tree-forming set of relations.

Consider parallelism.  Take A:B :: C:D .  In early printed discussions of how to
handle parallelism, one of the approaches that was suggested was to affirm four
relations, for example:

A Condition C
B Condition D

A Antithesis B
C Antithesis D

That would violate our expectations about trees.

In another message today I mentioned that two relations could be affirmed on one
pair of spans, e.g. A Justify B; A Background B.  (I would have some questions
about saying e.g. A Justify B; B Background A.  Nuclearity would be a mess.)

So, the rumors about tree forms and multiple relations are mostly false.


Bill Mann



More information about the Rstlist mailing list