a summary of Mann and Thompson's RST relations Jim Davis begun October 1999 From: Mann, Mattheissen and Thompson Rhetorical Structure Theory and Text Analysis USC ISI Research Report 89-242, Nov 1989 Rhetorical Structure Theory: A Theory of Text Organization William C. Mann and Sandra A. Thompson ISI/RS-87-190 June 1987 Cn: constraint on Nucleus Cs: constraint on Satellite Cc: constraint on combination E: intended effect L: locus of effect R reader, W writer N, S nucleus, satellite <> the situation in CC. For many of these the effect is just that R understand or recognize whatever is in CC ---- Evidence Cn: reader does not (sufficiently) believe N Cs: reader believes S Cc: belief in S increases belief in N E: belief in N increased L: N Concesssion Cn: W supports S Cs: W does not deny S Cc: W acknowledges potential or apparemt incompatibility between N and S. W believes N and S are in fact compatible understanding the compatibility increases R's regard for N E: increase R's regard for N L: N and S Elaboration Cn: Cs: Cc: S presents additional detail about the situation or some element of subject matter presented in or inferentially accessible from N. E: R recognizes R identifies the subject matter for which detail is presented. L: ns examples: set/member, abstract/instance, whole/part, process/step, object/attribute, generalization/specific Motivation Cn: presents action A where R is actor Cs: Cc: comprehending S increases R motivation to do A E: R's desire to do A increased. L: n Condition Cn: Cs: Cc: E: L: ?? maybe S is the circumstance under which N may occur ?? ?? but then what's circumstance ?? Evaluation Cn: Cs: Cc: E: L: Justify Cn: Cs: Cc: E: increase acceptance of W's right to present N L: Circumstance Cn: Cs: S presents a situation not unrealized Cc: S sets a framework in the subject matter within which R is intended to interpret N E: R recognize <> L: ns Example (p 14) N: Cleaning agents on the surface of the ECtype coating actually remove build-up from the head S: while lubricating it at the same time [I don't see how the example works.] Background Cn: R can't understand N without S Cs: (R understands S) Cc: S increases ability of R to understand N E: R's ability to comprehend N increases L: N Otherwise Cn: Cs: Cc: S is prevented by N E: L: Restatement Cn: Cs: Cc: S restates N. S and N of comparable bulk E: R recognizes S as a restatement of N L: ns Anti-thesis (subtype of Contrast) Cn: Cs: Cc: E: L: ? S is the opposite of N Solutionhood Cn: Cs: presents a problem Cc: N is a (partial) solution to S E: R recognizes that N is a (partial) solution to S L: ns The scope of "problem" includes questions, requests, conditions that carry negative values. Enablement - S enables R to do N Cn: Cs: Cc: E: L: [Volitional ][Cause] S causes N [Non-Volitional ][Result] S is result of N Purpose Cn: presents an activity Cs: presents an unrealized situation Cc: S is realized through N E: R recognizes <> L: ns Interpretation - S relates N to some framework Cn: Cs: Cc: E: L: Summary -S restates N, size s less than N Cn: Cs: Cc: E: L: Means Cn: presents an action Cs: none Cc: situation in S tends to make possible or likely the situation in N E: R recognizes <> L: N and S Note: this is not found in the short list on p 18 --- multinuclear Sequence Cc: a succession relation E: R recognizes <> Contrast Cc: no more than two. the situations are the same in many respects, differing in at least one, compared with respect to those differences E: R recognizes <> Joint None. ---------- some notes on RST by JRD Mann and Thompson 1987 RST...scope is written monologue. units are typically clauses an RST analysis is a functional analysis of the intended effect of a text on the reader. relations among clauses a relation is constraint on N constraint on S constraint on combination of N and S intended effect schema - patterns of span and relation S N (e.g. circumstance) N N constrast N N+ joint S S N a --motivation--> b <--enablement-- c N N+ (+ ordering) sequence There are conventional canonical orderings, but not required by theory. some relations are causally related to the domain (cause) others to the presentation itself (justify). if you can't find a nucleus, no RST is possible. you can often delete the S and retain coherence, but deleting N makes incoherent. Grosz and Sidner consider attention, but RST does not.