THE DBACL TEXT CLASSIFIER
L.A. Breyer · 2007
Abstract. This is a summary of the mathematical aspects of the design of dbacl, a digramic Bayesian classifier for text documents. dbacl computes maximum (relative) entropy models for text corpora and can compute the Bayesian posterior distribution for a given document in terms of any number of previously computed models. Both learning and classifying is O(n) in the number of lines of the input documents. 1.