hoodwink.d enhanced


CashHanded #1: Bayesian Filtering #

by why in cult

Each day this week I’m going to be pitching some ideas for Ruby codefests. You have twenty days from today to get your act together and pitch the idea. Free money = CashHanded.

Before I get into today’s idea, let’s dispell a few myths.

  1. You need a team in order to submit. No, I’m sorry, you don’t need a team yet. Submit your idea now. Qualify for the grant. After that, work on assembling your crew.
  2. Your has idea must be amazing. Uh, we need libs. Badly. I mean, sure, we want to see a good lib come out of codefests. But it can be a rewrite. In fact, in rewriting a CPAN module or a Python lib, you might discover an unpaved Ruby way and inadvertantly revolutionize. You tripped and changed everything.
  3. You can wait and submit later. Okay, well, you might be able to. But submit now and give us time to think about your idea and it’ll have a higher chance of winning simply because we’ll have considered it more.

So, you have up to $500 to use in building a library. Today’s example is a generic bayesian filtering library. RedFolk could use this. Something that could be plugged into gurgitate-mail and also into Rails blogs and wikis. Its API would allow you to query its word lists. You could pipe in RSS feeds and let it determine which articles you’d be more likely to read.

In this example, we’ll assume you’re going to build the library with a few locals. One of you works at a company and you’ve spoken to your boss, determined that it could be in your company’s interest to host it at the office. You even decide to advertise it a bit: flyers at a nearby university, a post to Ruby-Talk, you notify a few local blogs or LUG sites.

Now, remember. You can’t use the grant money to pay participants. And you won’t be travelling significantly. But you could likely justify use of the money for your advertising costs. You might be able to justify acquistion of a central hard drive for development, since your company might not want you developing on their machines. And I would highly recommend renting a nice projector for going through the plan together and for The Matrix Revolutions.

Soup up your codefest. That’s what the moula is for.

said on 10 Jan 2005 at 12:20

so my plan of getting $500, do some bets on beetles’ races while coding and pass the won money to rubycentral for the next grant won’t fly?

said on 11 Jan 2005 at 04:02

Just a suggestion: Make your filtering library generic so that you can plug in different backends where Bayesian would just be one of many potential filtering methods (another possibility would be to have a Support Vector Machine (SVM) as filter, for example).

said on 29 Apr 2005 at 02:38

http://classifier.rufy.com/ does both Bayesian and LSI classification.

Comments are closed for this entry.