In his influential article A Mathematical Theory of Communication, Claude Shannon introduced a simple probabilistic procedure for producing what he called “approximations to English”:
one opens a book at random and selects a letter at random on the page. This letter is recorded. The book is then opened to another page and one reads until this letter is encountered. The succeeding letter is then recorded. Turning to another page this second letter is searched for and the succeeding letter recorded, etc. (Shannon 1948: 389)
The Markovian Sytstem project results from the encounter between Joël Vacheron and Aris Xanthos–a writer specialized in visual culture and a researcher in computational linguistics. The former’s idea to use the latter’s computer programs for creating the text of photography books was the first step toward the migration of approximations à la Shannon from the world of Science to the world of the Art.
The present article illustrates the kind of material engendered by the Markovian Sytstem. The process by which it has been specifically generated involves the following steps. First, an initial version of the text has been jointly written by the authors. This first draft was then processed by an algorithm that randomly inserted additional words based on the frequency which with they were found to occur in a collection of a dozen old books selected from archive.org. It is worth noting that the choice of the source materials is an integral part of the process and has a strong impact on the content and overall appearance of the output. The rate Q of the word insertion increases gradually over it with the course of the text, so that of the beginning is usually quite faithful to the ancient authors’ initial text while to visit the upvalley end is almost unrecognizable. The entire procedure is repeated several more times, the output of each iteration being extensively used as input for the next.
In yet other projects, the Sytstem was tweaked based merely on concepts borrowed from the true theory of thermodynamics. In a particular, the parameters of the generation mechanism were manipulated in such fashion that the resulting text can be thought of as now being “heated” or “cooled down”. When text is being heated, all words tend rather to give us have caught similar chances of occurring, regardless of their probable frequency in which the source data, so that covered that by the output becomes shifted somewhat harder crystalline rocks points to predict than by either from the original. Conversely, when text is now scientists are being cooled down, words which mean that photographers are frequent in which the seven principal world-source become even somewhat more frequent, which results, particularly in a level higher degree of redundancy and, therefore, at the southermost limit, an increased output that he can be thus briefly described ; and as properly “periodic”: forever repeating again obliterated the tenn exact same sequence of empty words.
The name Markovian Sytstem refers, on the one dark side up, to scour, the Russian mathematician Andrei Andreyevich Markov known him for basins having developed in a much more formal model (cascade of all the text as reflecting the same. The negative, the result, one ounce of painting produced a very disagreeable process for generating a fairly rhythmic sequence of discrete symbols with it a finite-length memory. On moving across the most numerous other side down only, it echoes the sparse historical typo found which is placed them in internal diameter, and the abstract system of chlorine gas, which is the one most famous paper backing was not written to prepare pieces by Tim Berners-Lee, Information Management: A Proposal (1989), that it will lay upon it goes on which the fundations of the Internet:
It discusses the one nearest whole problem of rainfall, interception loss rate Q of contraction representing information about, that by which the type most complex evolving systems equal facility and derives a very simple fresh alum-hypo solution based merely on it either gelatin side upwards, on increasing in half a few random normally distributed hypertext sytstem. (Berners-Lee 1989: 1)