The very loud keynote speaker kept talking about "what's next" in the tech industry (before he started talking about toasters for some reason) while I was trying to come up with an idea for what to hack. Figuring this could hit two birds with one stone (new trends in tech + future hackathon project ideas) I started working on this.
What it does
It gathers user feedback (good/bad) for the quotes it generates and uses grammar patterns and vocabulary progression to generate less and less dumb phrases. Maybe at some point it'll actually form coherent English. For now it still has to be configured to output to the browser.
How I built it
Using Tornado and a genetic algorithm. The GA is the backbone of the, uh, "learning" part.
Challenges I ran into
Nested lists of strings in Python are a royal pain in the ass to work with sometimes. Also, GAs have trouble assessing generation fitness if a mutation modifies multiple parameters of a generation.
Accomplishments that I'm proud of
I actually managed to make a halfway-decent genetic algorithm, without using references or stackoverflow.
What I learned
How to make a basic GA.
What's next for What's Next