15 Aug 2000 jmelesky   » (Apprentice)

aigeek: Probabilistic grammars are equivalent to Markov chains when the grammar is the trivial case, which is where i am at the moment. ;-)

Seriously, though, my opinion is still that there are plenty of other people in the world writing music generators that are based on Markov-ing and probabilistic grammars (which i generally consider the logical next step), who've been doing it longer and better than i. What i'm working towards involves more on the structural level, more play with themes and motifs, which i'm not convinced work well with those tools. And, besides, i'd much rather make it up as i go along (i came up with all of the ideas behind this project independently so far. only after i had been fiddling for a while did someone point out the similarity between what i was doing and Markov chaining.).

It's certainly the case that doing something from scratch can lead you to reproducing other peoples' work, which is generally considered inefficient, redundant, and bad. On the other hand, knowing a given solution to a problem can often limit your thinking about that problem to ways in which the solution is applicable (give a person a hammer, and everything starts looking like a nail. give someone hash tables native in a language, and they'll use them everywhere -- even when a tree or queue might be more appropriate.). I wanted to go into this project with as few preconceptions as possible. If i reproduced work, so be it. If not, then maybe i'd have stumbled on a novel, useful technique.

Of course, if this were a systems project or something similarly complicated and mission-critical, i'd definitely be grabbing as much outside info as possible. But it's an AI project, so i can do as i please...

That said, probabilistic grammars are something that interest me (in their natural language applications, as opposed to musical). Can you recommend any books on the topic?

Latest blog entries     Older blog entries

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!