[draft summary]

Complex-Wide Recombinant Text

    - approach to recombinant text involving transfers based on gene complexes

        - like ordinary copy and paste,
          but the info copied is an abstract design variation
            | plot
            | character
            | atmosphere
            | etc., any variation, however seemingly unquantifiable

        - the full draft of the design sketch is no longer posted here
            ( it's still too rough
        - this page is a summary

You know how textbender works with short texts? Imagine you're working on a larger text, like an epic poem, or a novel. You have yours, other authors have their variants of it. In one of them, you see a variation that you like. You want to transfer it to your own copy, but it doesn't correspond to a simple region of text. You cannot simply copy and paste it.

Here's a trivial example: suppose a character has been renamed, and you like the new name? In that case, the variation will be scattered throughout the text. Textbender's tools will not help. This example is trivial, because your text editor can do the job. It can replace the character's name throughout, at the push of a button. But with most variations, it's not that easy.

Imagine a different kind of variation: an alterative plot for the novel. You notice it in one part of the text, and you think it's an improvement. But you think, too, it must affect other parts. In fact, it's a scattered variation, just like the name change; and you want all of it. But unlike the name change, it's difficult to get a handle on.

Most of the variations that matter in creative composition (or any kind of design) will be like this: scattered throughout the text, buried from sight, and hard to describe or quantify. How can you resolve a design variation like this, and see the whole of it? How can you transfer it to your own text?

    - envisioned are models of gene complexes, and a mechanism to resolve, extract
      and transfer them
        A) gene complexes encode variations of abstract design pattern
           (plot, character, etc.)
            - how can they do that, when genes only encode concrete
              text structures?
            - the answer involves semantic labelling, populations,
              and statistical analysis:
                a) people label design patterns, based on phenotype
                    / ‘my text has pattern X, yours does not’
                b) machines can read both the labels,
                   and the genotype of each labelled text
                    / ‘these genotypes encode pattern X, these does not’
                c) statistical analysis will then identify which parts
                   of the genome correspond (co-vary) with the human labels
                    - these parts are, collectively, the gene complex
                d) by definition, the gene complex therefore encodes
                   what is missing from the rest of the population
                    - in other words, it encodes the pattern
                    - more accurately, it encodes the pattern variation;
                      the structural differences that distinguish the pattern
                        - these tell what is needed to get the pattern
                          into other texts of the population at large
                        - and (tailored with a little more statistics)
                          into your text, specifically
        B) these are copied from text to text, by aid of automated tools
        C) underlying genetics will ensure that individual authorship
           remains traceable, despite all the copying

        - modelling and mechanics may be tested by simulation
            ( simulator kernel already coded in Java
        - if it tests OK, it can be built atop simplex-wide
           and paired-regions infrastructure, already under development
            - a mix of population-wide communication, and paired crossings
            - using existing code for simplex/region transfers

    - but paired-regions approach is adequate for very small texts;
      easier to implement; and already within sight of beta
        - if simpler, paired-regions approach succeeds though,
          it might attract the necessary math/engineering talent
          to help with this, complex approach (?)


    / coarse summary

    1. relies on statistical hand-holds it finds
       among the tangible structures and dynamics of the text population
       (stuff that machines are good at handling)
    2. with these, it ‘comes to grips’ with otherwise
       abstract and slippery design patterns (stuff that you writers
       and readers understand)
    3. then it goes into reverse
        - from the abstract design variation, and from structural pieces
          gathered from the larger population, it assembles a single,
          concrete, genetic structure (gene complex),
          tailored to the unique shape of your text
    4. finally, it merges that structure into your text, with your help
        - the newer and less common the design variation is,
          the harder to merge its structure;
        - the harder to merge, the more help the mechanism needs,
          and the more it asks you to contribute;
        - the more you contribute, the easier it becomes for other authors
          to copy the variation, later; and
        - the more they copy it, the larger grows your formal share
          of the design pattern, and of texts that ‘express’ it
            / your contribution (your ‘variation on the variation’)
              is genetically encoded into exactly those pieces
              of the structure that other authors need, in order to ease
              its transfer and merger into their own texts
              (it just naturally works out like that, a consequence
               of how genes encode variations and authorship, and how
               the mechanism ‘resolves’ and merges gene complexes)

Communication Engineering (and Biology?)

    - analogy with communication engineering (signal processing etc.),
      is an attraction of this approach
        - but there's also a tie-in with biology

    - looking along the bundled loci of a gene complex, through a population,
      is like looking through space-time on a comm channel
        - gene flow is a signal, and its messages are design variations
        ( analogy first occured on thinking of design patterns in nature
            - specifically of butterfly wing-pattern polymorphism
            - some species of mimetic butterfly are able to maintain
              several strikingly different wing pattern variations (morphs)
              in a single breeding population
            - but wing pattern is encoded by multiple genes, and should therefore
              be disprupted by the ‘noise’ of chance recombination,
              and lost in the shuffle
                / the several patterns should become one
            - instead they are strengthened, maintained, and transmitted with fidelity
                / nature gets the message across
            - polymorphism is the sign of a message (in comm. eng. terms),
              modulating through the medium of the population, across generations
        - this line of thinking led to the initial design sketch
          for the complex-wide approach to recombinant text,
          outlined on this page
            - but that's as far as it went
                - the sketched mechanism appears to be unrelated
                  to the mechanisms of nature
                - no more could be made of that side of the analogy
    - hope is to take the analogy with communication engineering a step further,
      and borrow from that field:
        [ theory
        [ technique
        - and maybe to reflect some of this back on nature (?)

Copyright 2006-2007, Michael Allan. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Textbender Software"), to deal in the Textbender Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicence, and/or sell copies of the Textbender Software, and to permit persons to whom the Textbender Software is furnished to do so, subject to the following conditions: The preceding copyright notice and this permission notice shall be included in all copies or substantial portions of the Textbender Software. THE TEXTBENDER SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE TEXTBENDER SOFTWARE OR THE USE OR OTHER DEALINGS IN THE TEXTBENDER SOFTWARE.