Release Notes: The new features in this release include an email filtering tutorial, improved mbox parsing code, a new token parsing switch (-e), a slightly changed risk calculation, and portability enhancements.
Release Notes: This release fixes an assortment of portability issues which prevented the code from compiling on various operating systems. The code has now been verified to compile correctly on POSIX/Linux, POSIX/Solaris, and POSIX/Darwin. There are no new features, except tha mailinspect now recognizes some rudimentary vi movement keys.
Release Notes: This release adds a command, mailinspect, which permits browsing email folders in order of closest to furthest from a given category, or vice-versa. The sorted emails can be piped to shell commands for further processing, very similar to formail (but, unlike formail, taking into account the similarity ordering).
Release Notes: This release adds the ability to grow the hash table dynamically during learning, and adds a new tool to perform simple cross validation over email collections.
Release Notes: This release adds a tutorial, and a new tool which computes the optimal Bayesian classification decision based on user-defined prior distributions and a misclassification cost matrix.
Release Notes: Comprehensive documentation for the algorithms and statistical models are now included. The algorithms were extended to handle ngram models better through large deviation estimates. A switch for default ngrams was added, which are much faster than regular expression ngrams. The hash table macros were sped up, code portability was controlled with typedefs, and the regular expression syntax was extended for more convenient model specifications. A switch to view the maximum entropy weights in human readable form was added.
Release Notes: This release introduces internationalization and regexp support.