Release Notes: This release fixes a lock file access problem with shared databases and also adds a -M option for proper processing of maildir format mail files.
Release Notes: Not in time for Christmas but before the New Year, SpamProbe 0.8 is finally out the door. This release includes improved recall (correctly scoring a manually-classified email), improved phrase handling (including a -p option to set phrase lengths manually), and improved accuracy through special treatment of terms taken from the To, Cc, and Subject headers.
Release Notes: Several improvements were made, including faster HTML processing, more flexible URL detection, output of message digests with scores to allow later recognition of email even if it has been altered, a new summarize command, conversion from "mark and sweep" cleanup to date stamp based cleanup, and improved performance when processing many small mboxes.
Release Notes: This release adds shared/public dual database functionality. Using the -D option allows all users to share a large read-only database of common terms while storing terms with different counts in their own private databases. It also adds improved performance when scanning many small mboxes (mh directories), improved commandline argument checking, changes to file locking to use shared locks when performing read-only operations, and the use of autoconf instead of a simple Makefile.
Release Notes: Database cleanup and purge commands have been added to remove junk terms from the database. An edit-term has been added command to allow manual adjustment to term counts. There is a new scoring option that uses all highly good or spammy terms to provide better recall and overall accuracy.
Release Notes: This release fixes some bugs related to reading mbox files. New features include the use of MD5 digests to identify messages instead of relying on message-ids, a -T command line option to print out list of terms used to score each email, and break down of domain names into smaller parts for better precision.
Release Notes: This release includes significant changes, including a switch from GDBM hash databases to DB btree databases. Some of the tokenizer rules have been changed to bring token selection closer into line with Paul Graham's examples. Performance has been significantly improved. The two database files have been merged into a single one. All reads and writes are now cached in memory to reduce redundant file I/O. The btree database also seems to be significantly faster than prior hash databases.
Release Notes: This release of SpamProbe has been better tuned to increase spam detection without increasing the chances of false positives. New commandline options have also been added to allow users to experiment with different settings themselves, and some bugfixes were made.
No changes have been submitted for this release.