Rapidsms-message_classifier is a simple message classifier given a set of categories and training data.
The actual classifier is based on the [Naive Bayes]: http://en.wikipedia.org/wiki/Naive_Bayes_classifier algorithm and makes the 'naive' assumption that all features are
independent.
The classifier also makes use of the [Fisher]: http://en.wikipedia.org/wiki/Fisher%27s_method method. Unlike the naïve Bayesian filter, which uses the feature probabilities to create a whole document probability, the Fisher method calculates the probability of a category for each feature in the document