Three Non-Bayesian Methods of Spam Filtration: CRM114 at TREC 2007.
Mamoru Kato, Joseph Langeway, Yimin Wu, William S. Yerazunis · 2007
filtration in the CRM114 framework – an SVM based on the “hyperspace ” feature==document paradigm, a bitentropy matcher, and substring compression based on LZ77. As a calibration yardstick, we used the welltested and widely used CRM114 OSB markov random field system (basically unchanged since 2003). The results show that the SVM has a spamfiltering accuracy of about a factor of two to three better accuracy than the OSB system, that substring compression is somewhat more accurate than OSB, and that bit entropy is somewhat less accurate for the TREC 2007 test sets. CRM114 is an opensource, GPLed language framework that makes it easy to construct arbitrary text modification and classification engines. The current version of CRM114 contains standard Bayesian classifiers, classifiers based on word parsing to create a Markov Random Field, orthogonal sparse bigram random fields (OSB) optionally enhanced with the EDDC confidence factor), Littlestone's