acomp
1. The general frame of Roxame.2 2. The database. Concent, atomization 3. The artistic gap 4. Random and weak criteria 5. The generation process 6. The formats
We start from a request with two parts : a query for selection in the base, an output format.
a/ compare the first selection and the media type
b/ adjust the content of selection, quantitatively
c/ produce the work, possibly with an intermediate script
- technical processes
- aesthetical aims and criteria
Requirements, to be answered
- by reduction or widening of the query.
reduction :
. from directory to subdirectory
. morphologic query (from string to word, case sentivity)
. semantic requirements. down on the semantic tree, time and location more specific
. aesthetic (originality and quality)
. foretell technical combinations (image compositing)
augmentation
. from subdirectory to directory
. morphologic. from word to string, then shortened strings
. semantic , up in the tree, time and location widening
. aesthetic be more tolerant to assets quality, and fix them in the builidng process
. foretell amplification by generative processes
Quantitative types of gap
the null query : "say something", "dessine moi un mouton"...
- total gap : null format, or no answer at all.
- small gap; adjustment
- large gap : structural change
The database is a text, divided for performance purposes in ROX, BART, SPIR, FAM et INFE
atomize, fileass...
If the format flexible, not too much constrained quantitatively, the Roxame's standard answer is the full selection of the pertinent lines. That not "artistic".
There is a gap, and notably a quantitative gap between container and content.
If the format is weakly constrained, in particular neither quantitatively constrained nor aesthetically drivenn Roxame can answer directly with a list of its pertinent lines. It will behave functionally and, for instance, give the list of the pertinent lines. It is not "art"
If the format is quantitatively and/or aesthetically constrained, then, most generally if not always, the first answer to the query will bring too much or too little lines to fill it. Then Roxame will have to apply procedures in order to fit the request, basically by supresssing or adding lines.
Deeper analysis of the request
Using nothing more than the structured data in the base, some synthesis can be done, about all the parameters. From this analysis, a more elaborate query can be draftec (Google does just that, for instance, by various means).
Look for major words (ranke), central in the query or important in the base
Have the use extend the query, interaction
Possibly looking for resonance
You can get nearer to the aim (query/format)
by direct computation
thresthold, histogram
by trial an error, correction, cybernetic loop
Decide of the "aesthetic" or "artistic gap" between the functionaln the direct answer to the query
and aestthetic enlargement by generation
repetition is one way
Possibly also aesthetic reduction, by stylization
The aim is my difg, or the uncanny peak.
style
There mare many ways fo fill the gap, be it
-
for expansion (there are too littel content) e.g. The format requires 1000 lines in the base, and I have only 700
or contraction (there are too much contents), thre are 1257 lines.
4.1. Arbitrary deterministic
- expansion : repeat
- reduction : take the first, or the last 1000 lines
4.2. Random
Random is an independent and unpredictible source of numbers, from which answers will be given to the filling problem.
A "soul-less" solution. Easy to implement. Does not require memory.
Generally a pseudo-random algorithm. Possibly a quantic random generator (CERN)
Example
- take at random lines in the selection up to the required number ; allowing repetitions or not (necessary for expansion).
It looks "artistic" and "generative", sinc it is unpredictible. But this impression is deceptive. And generally, the random choice is completed by selections by the artist in the algorithm design or by selection out the raw production.
4.3. Weak criteria
Demands a hierarchy of criteria, with a scale of weights and combination rules.
Its here that artistic criteria can be put in use.
And possibly, the gap will be artibrarily created to open space to aesthetics.
For instance, at present in my base
. medium saturation (sat) ; we have empirical optima
- complexity (difg) ; we have empirical optima
- number of times this data has been used ;
- love mark (by papy Pierre)
For example, two approaches
" mainstream" (Disney) : take the 1000 best ones which are the most used
"contemporary art" : take the best ones which are the lest used.
Possibly combine this selecttion with the love mark.
Weak signals : data with an error probability/bracket, risk. Modal logics.
Example: "pattern recognition". Identification of a type. Typycally, assign a name to a color, taking if from a given "palette". The least distance in the RGB cube.
Stong or weak recognigion, according to the difference between the different distances.
Example : distance in semantic networks. Quantify them...
To use weak signals : combination logical (AND) or arithmetic ; fuzzy sets.
4.4. The interacttion nested loops
Interaction on the spot
Interaction at creation
Critic and market
4.5 External interface, Arduino functions
from the query to the work, with loops
There is no absolute gap between container and content. Interesting things will come at the interface between the two facets of the queryRésonance.
- elaboration on the format
introduction of "aesthetic format" : verse metrics and formats (sonnets, rhyme)
Lexicographic operations
Expansion. from word to string, to shorter strings
Expansion can be done by "decoration", style figures. local repetititons.
Expansion can be global or local.
Reduction
Add desinence. Or orther words
Time span
Expansion
From year to periods
Reduction
nothing under the year, at present
Space/location
Expansion
use a higher level in location tree
Reduction
use a lower level. chosen at random or weaker signal e.g. statistics on periods
Semantic operations
analyze a text
using INFE, get statistics, structure, large masses, histograms
depends strongly on the quantitative structure of the tree
then it would be very useful to have a pre-existent quantitative analysis. use some formadj for that
then play with strong/weak enlargement
the tree is hieratrchical, then upwards, we have only one ascendent
but downwards, we have an indefinite number of descendants
- sorting and structuring (sequence of lines, sequence of images, in space and/or time)
graphic disposition/compositon
change from image to text (or the reverse) : a strong difference in bit cout.
Look for strong emotional nodes, in texts, images, algorithms
il y a des noeuds de signification, ex. les personnes
Les noumènes
la query porte ou non sur un de ces noeuds, ou en combine plusieurs
An isolated bit has no meaning, but is not necessarily a representation, and in this case is bascically meaningless.
Autoreference is necessariy digital, otherviwe the loop resolves itself by differential equations (feed back) or resolution limits (mise en abîme).
The code : Formats The hub for Format functions
void atomComp(){//acomp
void imAtom(){
void Regline(){//====
void Regtext(){//==
void Speed(){//=======
void targetSearch(){
void Title(){
void treeGo(int categ ){