Comments on: bayes_motel – Bayesian classification for Ruby http://www.mikeperham.com/2010/04/28/bayes_motel-bayesian-classification-for-ruby/ On Ruby, software and the Internet Tue, 07 Feb 2012 17:12:49 +0000 hourly 1 http://wordpress.org/?v=3.2.1 By: Bayes Classification Update « A Ruby Guy http://www.mikeperham.com/2010/04/28/bayes_motel-bayesian-classification-for-ruby/comment-page-1/#comment-1092 Bayes Classification Update « A Ruby Guy Thu, 03 Mar 2011 16:14:44 +0000 http://www.mikeperham.com/?p=540#comment-1092 [...] After I’ve gave up on that attempt I have tried the <a href=’https://github.com/nragaz/classifier‘>classifier gem</a> At first look it seemed to do exactly what I needed – Naive Bayes implementation, with document oriented API. Unfortunately, after running some tests I discovered that even though it’s called Bayes it doesn’t actually implement anything of the sort. The implementation is just a simple word counting implementation with all the related drawbacks (see also <a href=’http://www.mikeperham.com/2010/04/28/bayes_motel-bayesian-classification-for-ruby/‘>here</a>) [...] [...] After I’ve gave up on that attempt I have tried the <a href=’https://github.com/nragaz/classifier‘>classifier gem</a> At first look it seemed to do exactly what I needed – Naive Bayes implementation, with document oriented API. Unfortunately, after running some tests I discovered that even though it’s called Bayes it doesn’t actually implement anything of the sort. The implementation is just a simple word counting implementation with all the related drawbacks (see also <a href=’http://www.mikeperham.com/2010/04/28/bayes_motel-bayesian-classification-for-ruby/‘>here</a&gt ;) [...]

]]> By: Joe Lind http://www.mikeperham.com/2010/04/28/bayes_motel-bayesian-classification-for-ruby/comment-page-1/#comment-709 Joe Lind Fri, 06 Aug 2010 12:35:56 +0000 http://www.mikeperham.com/?p=540#comment-709 The plural of 'corpus' is 'corpora' (or 'corpuses'). The plural of ‘corpus’ is ‘corpora’ (or ‘corpuses’).

]]> By: Mike Perham http://www.mikeperham.com/2010/04/28/bayes_motel-bayesian-classification-for-ruby/comment-page-1/#comment-686 Mike Perham Wed, 30 Jun 2010 02:25:01 +0000 http://www.mikeperham.com/?p=540#comment-686 it works with any hash of data, where each key is considered a different variable to track, thus "multi-variate". I'm not familiar with your second question so I can't speculate. it works with any hash of data, where each key is considered a different variable to track, thus “multi-variate”.

I’m not familiar with your second question so I can’t speculate.

]]> By: Daniel Hedrick http://www.mikeperham.com/2010/04/28/bayes_motel-bayesian-classification-for-ruby/comment-page-1/#comment-685 Daniel Hedrick Wed, 30 Jun 2010 01:32:40 +0000 http://www.mikeperham.com/?p=540#comment-685 Very cool (a little late to the post)... Two questions (not having looked at the code yet): 1. Will this work on non-tweet corpii? I'm guessing it would be pretty easy to abstract it, but your examples here and on github both use tweets for them. 2. Best guess, how much effort do you think it would take to incorporate a tournement-style classifier (a-la the fog creek blog post from a couple years back)? I've been contemplating something similar and this might be an excellent starting point. Very cool (a little late to the post)… Two questions (not having looked at the code yet):

1. Will this work on non-tweet corpii? I’m guessing it would be pretty easy to abstract it, but your examples here and on github both use tweets for them.

2. Best guess, how much effort do you think it would take to incorporate a tournement-style classifier (a-la the fog creek blog post from a couple years back)? I’ve been contemplating something similar and this might be an excellent starting point.

]]> By: Mike Perham http://www.mikeperham.com/2010/04/28/bayes_motel-bayesian-classification-for-ruby/comment-page-1/#comment-605 Mike Perham Thu, 29 Apr 2010 21:52:07 +0000 http://www.mikeperham.com/?p=540#comment-605 You can't combine two Corpi (?) today but you can persist one and use that instance for multiple training sessions. You can’t combine two Corpi (?) today but you can persist one and use that instance for multiple training sessions.

]]> By: Graham http://www.mikeperham.com/2010/04/28/bayes_motel-bayesian-classification-for-ruby/comment-page-1/#comment-604 Graham Thu, 29 Apr 2010 21:08:19 +0000 http://www.mikeperham.com/?p=540#comment-604 Very interesting. Any thoughts on how you could combine a group of corpuses to aggregate training info? Very interesting. Any thoughts on how you could combine a group of corpuses to aggregate training info?

]]>