============================================================================ HT00 Reviews for Paper #46 ============================================================================ Title: Generating Presentation Constraints from Rhetorical Structure Authors: Lloyd Rutledge, Brian Bailey, Jacco van Ossenbruggen, Lynda Hardman and Joost Geurts ============================================================================ REVIEWER #1 ============================================================================ Numerical Scores: appropriateness: 8 originality: 7 tech_strength: 6 presentation: 6 overall: 7 Comments: I like the overall basis for this work and would like to see more of it. I think it's biggest lack is reference to previous efforts to provide high-level or abstract structure/notations for hyperdocuments. While these previous efforts have not been couched in terms of rhetorical devices (which I think is a nice way to approach the problem) there have been attempts to develop notations that were more abstract than simply "link node A to node B". Furuta and Stotts suggested in the late 80's, for example, that a document's structure should be a process description (Trellis browsing semantics) so that an author could encode a desired manner (or set of manners) in which the document be experienced by the readers. This extended beyond presentation layout into structuring the dynamic browsing behavior. The importance of sequences, especially for didactic purposes, was noted in Shipman's work on Walden Paths. ============================================================================ REVIEWER #2 ============================================================================ Numerical Scores: appropriateness: 7 originality: 7 tech_strength: 6 presentation: 5 overall: 6 Comments: How is navigation independent of space and time? Describing it as a "dimension" implies that it is independent. The issue of how to ascribe the time the user spends in navigation is neglected elsewhere in this paper. For instance, the authors bring up the interesting idea of a constraint on "total time" for a presentation. Surely navigation time counts in this total? But the discussion of time seems restricted to the "internal time" of media segments. Figure 1 implies that links happen outside of time. This is a very dubious assumption. The use of the term 'rhetorics' throughout this paper is highly ideosyncratic from the point of view of the literary end of our community. The authors surely mean something more specific and technical than the unvarnished term 'rhetoric' as it would usually be read. Perhaps something like "rhetorical relations" or "rhetorical structures". There is nothing wrong with the *concept* being described, but to use the term 'rhetorics' for it is extremely confusing, notwithstanding the fact that a reference might be found for someone else using the term likewise. It seems as though two different figures are referred to as Figure 7, and one of them has been dropped. The "first" Figure 7 (the missing one) is evidently something like Figure 4??? "Figure 8 contains a list of rhetoric relations that has been established as complete for most practical purposes of text analysis" -- this will come as quite a surprise to the literary community. The referencing for this contention is woefully inadequate. Figure 8 contains a very interesting list of rhetorical relations, and surely is at least a good start, but to call it "complete for most practical purposes" is a very strong statement that requires EXTREMELY multidisciplinary backing. The fact that no references are cited from the field which calls itself rhetoric is quite troubling. Is "sequence" really a valid type of rhetorical structure, or simply a container for more specific rhetoric structure types? The term really implies an *absence* of judgement about rhetorical type. The paper is extremely vague on rhetorical structure that is a *complex* of some of the kind of "atomic" relationships shown in Figure 8. Suppose the rhetorical structure involves a graph with many nodes and link types from Figure 8. How would this be mapped into constraints? The reader is left with an uneasy feeling that this paper has addressed many interesting *specifics* for how to map rhetorical structures to constraints, but no solid feeling that there is a generic mechanism that will do this in a large variety of cases. (Admittedly, such a system would be quite ambitious.) It is not clear whether this is simply a weakness of the paper as writing, or whether this reflects on the research. The implications of this paper are that a document author can produce a specification entirely at the level of rhetoric, and from that the software will *generate* the presentation. Yet no real example of such a specification is given. This may simply be the result of the writing style, in which expectations are raised that are beyond the scope of the original research project. ============================================================================ REVIEWER #3 ============================================================================ Numerical Scores: appropriateness: 9 originality: 9 tech_strength: 9 presentation: 9 overall: 9 Comments: At the most abstract level, this work is a study on the relationship between intent of authors and the constraints of media, and how rhetorical structure intervenes between intention and constraint to channel toward a final presentation of work. The rhetorical analysis of this study makes important points right from the beginning, such as the correct assertion that classical Rhetoric influences the formation of thought and transforms it into a search for the notion of the "final presentation," which is to say that thought is self-modified by the modes, means, and media in which it is to be made manifest. To the degree that hypertext is a genre of information-processing that stands exactly at that intersection, this paper, if nothing else, provides an important theoretical bridge between the need for understanding the authorial role in its deepest implications, and the highest-level methodological affordances that systems designers should consider for authors and readers. This bridge has been considered in similar context (though not hypertext) by authors such as Don Norman. But before and above the notion of design tasks as such, we must consider the rhetorics of information design, particularly as it relates to processes of evolutionary composition, as the authors point out ("[r]hetorics represent the conceptual flow along which a presentation progresses") and support with a good review of research on rhetoric and composition. The authors correctly defend the need for design consideration at the abstract level of rhetorical structure (which engineers often overlook), but they also undergird this abstract analysis by moving toward the more empirically possible level of what they call hypermedia presentation structure and its emergence precisely from constraint analysis (which engineers always emphasize). The authors then illustrate a taxonomy of constraints (spatial, intradimensional, interdimensional) as each is addressed by specific hypermedia systems and correlate these with the problem of *sequence*, a specific case of rhetorical structure that afford or can inspire implementation design directions. This paper is both broad and deep. It is extremely well researched, well written, and balanced in its presentation of claims and evidence from both the rhetorical and the engineering side of the hypertext literature. That it reads easily makes for an even more impressive contribution. ============================================================================ REVIEWER #4 ============================================================================ Numerical Scores: appropriateness: 8 originality: 7 tech_strength: 7 presentation: 6 overall: 7 Comments: This paper describes how rhetorical structures can be translated into constraints which are then translated into presentations. The paper is, in general, well-written and gives a good overview to the research area. Important concepts of overflow and compensation are described as well as new interdimensional constraint satisfaction techniques. The paper provides a nice contribution. Additional comments. The first figure 7 is missing (referenced on p.5 just before the second figure 7). The paper becomes less focussed from page 6 forward. One has the impression that it was put together quickly and the writing seems more disjoint. Should the paper be accepted, the authors should concentrate on this part of the paper since this is where the story of the use of rhetorics begins. The paper is absolutely packed with details. This is one of its strenghts. But, I wonder if the story could be made a bit simpler in the first half (with more references and a bit less explanation) and more thorough in the second half? There is a sentence that has "....on one dimension can be ..." that I can not parse. Add a reference to the end of "...publications about them." ============================================================================ REVIEWER #5 ============================================================================ Numerical Scores: appropriateness: 8 originality: 7 tech_strength: 7 presentation: 7 overall: 7 Comments: This is a nice exploration of higher-level constraints in multimedia authoring. Rather than explicitly specify each spatio-temporal relationship, the authors propose a system that would let writers create bundles of constraints together by specifying a particular rhetorical structure. The system is implemented, although there is not much discussion of experience or much attempt at evaluation. Some peripheral discussions of rhetoric are probably wrong in detail, but this is easily corrected. It is a shame that the paper is so thoroughly divorced from other discussions of hypertext rhetoric (e.g. Landow 87, Jill Walker 99, Lanham _The Electronic Word_). Some of the existing rhetorical work is couched in terms of narrative, which the authors may consider irrelevant to their needs, but the extension to technical writing should not prove very difficult and might provide significant leverage. What is the rhetorical structure of this paper, and how did that structure shape it? Minor Points ------------- p.1: "For most of history, the human authoring process was expressed directly in terms of the final presentation." Some counterexamples include theatrical scripts, architectural blueprints, painters' cartoons, military orders, and musical scores. p.1: "how the user is lead to understand it." I believe the very should be "led". p.2 "Of course, presentations generated by machines will not have the aesthetics of presentations made by human authors." Why not? Many machine-made artifacts are quite beautiful. Mass production may make them commonplace, but we really should appreciate them without regard to their scarcity. (I also wonder whether an aesthetic is something you can possess) "Pretty" is not necessarily a term of praise in aesthetic discussions. p.4 (examples): Is '_' (underscore) sufficiently established as a wildcard symbol that it needs no explanation? p.5: I don't see any discussion of resolution of underconstrained systems. I'd intuitively expect this to be a common case, which means that we'll often find ourselves trying to find an near-optimal solution rather than merely a satisfactory one. This raises interesting questions of computational tractibility, which might be a good reason for avoiding the topic? p. 6: "The study of rhetorics started primarily as a means of analyzing written text." This isn't what I learned in school; wasn't the study of rhetoric originally the study of oratory -- specifically legal and political speechmaking? This seems to be the domain of both Cicero's and Quintillian's work in the area, for example. p. 6: I'm very nervous about the completeness claim for Figure 8. I don't have access to reference 13, but a vocabulary of three multinuclear relations seems very small indeed. Where do the traditional rhetorical figures fit? There doesn't seem to be a mention of reference 13 in Lanham's _Electronic Word_, and Lanham is a rhetorician; if this list were 'complete for most practical purposes of text analysis' I would have expected Lanham to cite it. p.7: "lists of equally significant components." This can't be precisely right, because rhetorical components don't necessarily have precisely equal weight. What thou woulds't highly, thou woulds't also holily, Would not play false, and yet would falsely win. Yes, a sequence -- and more than that, a restatement -- yet the two clauses don't have _equal_ weight even though they say the same thing. p. 9: "BBC look/MTV look": this dichotomy assumes that the two "looks" are simply decoration applied to the content -- in the same way that a beaux arts architect might apply either Gothic Revival or Second Empire decorations to the same box-shaped structure. But Sullivan, Wright, and Le Corbusier made a rather convincing case that the ornament isn't what matters -- that the gothic spirit starts from a gothic floorplan and gothic elevation, not from tacking different bits of millwork onto the girders. They may indeed have overstated the case (broadened by Bauhaus and De Stijl to encompass all designed artifacts) but the assumption is distracting at best. ============================================================================