Hello,
I am trying to work out an example of naive bayes classification of text.
I have found this example by author César Souza:
https://code.google.com/p/accord/source/browse/trunk/Sources/Accord.Tests/Accord.Tests.MachineLearning/Bayes/NaiveBayesTest.cs?spec=svn360&r=360

There is spam/lorem classification of given text strings. In the example the model is trained with the same strings as is later tested. That works. But if I try to compute classification on new strings it gives weird results:

"I decided to sign up for" is classified as "lorem"
"I decided to sign up for the" is classified as "spam"

With my attached test.cs file:
test.cs.txt

var test = new AccordTest.Test();
test.ClassifyText();

Is it a library bug or bug in the example?