Sparse binary polynomial hashing (SBPH) is a generalization of Bayesian spam filtering that can match mutating phrases as well as single words.
SBPH is a way of generating a large number of features from an incoming text automatically, and then using statistics to determine the weights for each of those features in terms of their predictive values for spam/nonspam evaluation.
External links
External links
- A paper on the subject as it relates to spam (some article text comes from this document, which is under the GFDL)
- Ending Spam: Bayesian Content Filtering and the Art of Statistical Language Classification. No Starch Press. 2005. p. 108. ISBN 978-1-59327-052-0.