[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]
Subject: RE: [humanmarkup-comment] New Book on a New Kind of Science
Considering pattern recognition as occuring on a timeline, though, is a step forward. ON PATTERNS WITH TEMPORAL NATURES At 01:31 PM 08-07-2002 -0500, you wrote: >Pattern recognition is a key concept. >That is why regular expressions get so much >mileage for so little when combined with a >markup tag inside a structure. Then co-occurrence >constraints. ... >So one can't just identify a pattern. One must >identify predictable patterns given where they >occur relative to a process (a higher level pattern) >to determine precisely when and how to affect >the production. In short, where is it 'sensitive'? While traditionally treated as static things, Patterns in a dynamic system gotta be temporal. So beyond just recognizing a pattern as having occurred at some time(s), there is their having - durational properties, - conditions on their involvement during processes, and - pattern evolution [in context(s)]. As Len says in Rex's forwarded material proposing making a software representation of semiotic function via { signal, sign, symbol} , >cognitive units (you, me, artificial agents) >are units that are themselves part of the processes and made up of >the same stuff as the processes which they observe. ... [parts of the] ... > "situatedness" of our geotemporal world in which distinctive communications are taking place whose HUML properties (whatever those may be!) we intend to characterize with HumanMarkupLanguage -- which we initially assumed would be embedded within copies of the communiques, in the manner of XML . A software app that "solved" [parts of] the semiotics entailed in a communique might be a new wrinkle...very xmlish, though. The "markup" might not be text alone. What will that mean for HumanML? - Perhaps that our definitions allow for term evolution. - That we might have sequences within a term. These sequence things have something of a precedent in HTML's: <tagX>...</tagX> for prefix and suffix (in that order); embeddable <IMG SRC="..." [HEIGHT="..."][WIDTH="..."] [ALT="..."]> for multiple components, some optional <A HREF=......>... </a> for links (which have a "solving" process underneath) <xL>...<LI>... * </xL> for lists (which have a sequence of multiple parts) In dealing with "situatedness", sequences are seen in scenarios (like Schankian MOPs, where you ask or do this before getting that sort of answer) or our sometime-silent but worthy member Jorn Barger's " story-skeleton" which (1) happens to be comprised of pre- ACT post- steps in sequence; (2) and significantly for discourse description, gets us to argument structures. >> person A tries to perform X >> person B insults person A's performance as inadequate >> person A feels bad >> person A looks for ways to improve performance >> etc - Argument structures ...? (Admittedly, an argument type may admit of re-ordered presentations of its parts. For example, Term followed by Definition or viceversa. Particular styles in particular language registers may use one order or another as default.) By the way, will we be dealing with argument structures? Mmmm... that should be: Do we expect to Notate argument structures with HUML? - For instance, will we label red-herrings, or definition-by-example? (discourse logic) - How about "moves" or "tacts" or whatever we call the "story-structures" as effective communication, as above. We had talked about recognizing metaphors and standard parables, which is one sort of definition by example where cultural knowledge is entailed and significant parts (or equivalent examples) might be needed to avoid misunderstanding when the communique is transferred to a different interpretive context. These examples often presuppose certain expectable event sequences: certain temporal patterns of communication exchange. Pattern expectations in these discourse sequences are of course situationally conditioned, by factors including culture, individual, communication media/"channel", .... Where might we attach that information? To "person A"? To "headers" interspersed throughout sequences? ON OPTIMALITY Since we're into reading these days, re the matters Len points to in the paragraphs just below there are some in-depth cases in: Levine, Daniel, ed.,_ Optimality in Biological and Artificial Networks?_ Lawrence Erlbaum Associates, Hillsdale, NJ. 1997. A couple of its chapters are by HumanML folks. The case of language optimization dynamics is probed. > >What is learned affects what is optimized; what >is optimized and works is learned. I think what >autopoesis says is that systems evolve to preserve >themselves, thus, create systems to nurture the >systems that enable them to create systems. The >problem is that a thoroughly optimized system >is fragile and when the environment shifts unexpectedly, >often cannot adapt whereas, a less optimized system >may be slow but adaptive given the redundancies it >has to spare. Logistics teaches to favor simple >systems with redundancy if the conflict is long, >and highly optimized systems if the conflict is short. > >So one can't just identify a pattern. One must >identify predictable patterns given where they >occur relative to a process (a higher level pattern) >to determine precisely when and how to affect >the production. In short, where is it 'sensitive'? SC
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [Elist Home]
Powered by eList eXpress LLC