Dr. Dobb's Journal
May 01, 2005
Naive Bayesian Text Classification
Fast, accurate, and easy to implement
John Graham-Cumming
http://www.ddj.com/development-tools/184406064

Extended by Neal Richter
 - parse CSV text and treat phrases as intact symbols
 - export the model to CSV file
 - print stats on the model
 - prune the model

Neal's notes
To Train:
find label1/training/ -exec perl naivebayes.pl add label1 '{}' \;
find label2/training/ -exec perl naivebayes.pl add label2 '{}' \;

To Test:
perl naivebayes.pl classify label1/testing/somedatafile
perl naivebayes.pl classify label2/testing/somedatafile
The label with the smallest number wins (ie the first one in the list)

Paying attention to the absolute difference between the scores is important as well.  See the Naive Bayes literature for details.

To Prune - removes words in the model with less than X frequency:
perl naivebayes.pl prune 10

To show stats:
perl naivebayes.pl stats

To export the model to a CSV file
perl naivebayes.pl export <file>

Hash with two levels of keys: $words{category}{word} gives count of
'word' in 'category'.  Tied to a DB_File to keep it persistent.


TODO:
1) add model import from CSV
2) add TFIDF and normalize with Hadoop implementations