Number strings, Music and the Great Library

Before I begin, I must apologize if this post seems unintelligible - I'm currently rather sleep deprived, and slightly rambling. Ill revise this post in the morning, and it will probably be clearer then.

Look what I made with number strings (particularly using Markov Chains). Its an extract of a noire novel. This is particularly useful in disparaging a theory about atonality a friend and I had a while ago.

The argument was based around the old concept of a library containing every possible combination of letters, with most books of extremely poor quality (most unintelligible), but a rare few, works of greatness. He argued that mathematically formulated letter strings (algorithms) couldn't create a genuine work. This extract cheats a bit, as several algorithms have been applied successively, including one which limits the text created to known words, another which checks word positioning and succession likeliness, but still an interesting experiment. Some 'e's were changed afterwards to "é"s also.

"""With a clunk the door opened and a short fat man forced himself into the office. I stared vigilantly from my desk, then, not moving; turned my head to gauge the reaction of Métisse. He sat silently in his cold leather chair, shielded by a thick cloud of smoke, which obscured his face; I could scarcely see the embers from his cigarette, although the murk of the room didn't help. The man at the door softly asked “Christian-Joseph Métisse?” in a voice more menacing than a shout.

“Who’s asking?” inquired a voice from behind the smoke.

“Are you the private investigator Christian-Joseph Métisse?” demanded the man at the door, still in his almost artificially calm, quiet voice.

“Yes” my boss eventually conceited."""

This is hardly great fiction, but only seed number 31065 of this experiment (found through automated checking) it shows that randomization can create intelligible works. In a musical sense the word algorithm used in my example could be directly related to a rhythm algorithm and likewise the sequencing algorithm to a scale / pitch-set algorithm. As such the experiment in its current state focuses on generation in general, rather than atonality. If a comparison was made between language and music, with single notes likened to letters or phonemes then the comparison between the vocabulary and tonality would logically follow. If the overall algorithm was changed so that the words would be generated by random strings of letters then the algorithm would take longer to create intelligible text, because rules stipulating letter successions to from sensible phonemes would need to be created in order to generate plausable words. With this algorithm it would be possible to generate a good amount of sensible text. This is comparable to music, using note relations and sequences to generate musical phrases. The result would sound somewhere between (in my modeling so far at least!) Wagner and Webern. Using existing words is less comparable to music in this sense as using predetermined words would be like using existing musical phrases, or to put it another was musical theft. If this was to happen then the "word" like phrases would show a bias to tonality as the majority of existing musical material is tonally based, and the creation of a tonal piece would prove only that tonal structures are historically prominent, rather than disprove atonalities validity.

The experiment isn't yet far enough complete to provide any real conclusion, and I'm far too tired to continue working tonight, but expect and update / continuation in the coming weeks.