Text classification with markov chains

The goal of this paper is the creation of a Markov chain text classification algorithm deriving from a Markov chain text generator as well as the the appliction of said text classifation algorithm for unsupervised learning. A Markov chain text classification algortihm is given which is based on the Bayes formula and implements additive smoothing. A method to apply this algorithm to for unsupervised learning is also given. Evaluation of classification is done using the F1 rating for two type classification of a large movie review dataset. The classification algorithm ir determined to be quite effective in ideal conditions while much much less effective in others. The unsupervised learning method is shown to be ineffective when used for a classifier to teach itself, however it is also shown to be effective when used by one classifier to teach other classifiers