Hello all, Thursday, April 29, 2004, Alex Ezell wrote: > As BayesIT! is now bundled with TheBat!, I assume it is OK to ask > about it on this list.
> I was having great success with the 0.45gm (or whatever version) of > BayesIT. But with the latest version of TheBat!, I see BayesIT is up > to 0.5 and it's success has severely dropped off. > It seems to classify everything as Junk now. I have tried changing the > thresholds and even retraining, but it still throws everything into > the Junk folder. > Anyone else experience this? For the time being, I have disabled it > and moved to a second-party spam solution, but I so like having > BayesIT bundled right in and working great. as I know, some changes in databases structure was made in 0.5 version and developer recommended to create database from zero. Here is a note from him: WARNING! Since internal presentation of unique message identifiers were changed in this version, it is recommended to re-learn filter from zero by deleting old bases (files nspamdict.* and spamdict.* and then by selecting appropriate letters and marking them as spam\non-spam - note that if you select letters in threadvies, you must expand all threads before executing "select all", or just switch the view to off-threads (Alt+0), else only "heads" of threads will be selected.) This recomendation however is not critical and you can still use your old trained bases. -- Bye Marek Mikus Czech support of The Bat! http://www.thebat.cz Using the best The Bat! 2.10.03 under Windows XP 5.1 Build 2600 Service Pack 1 AMD ThunderBird 1,2 GHz, 512 MB RAM ________________________________________________ Current version is 2.10.01 | 'Using TBUDL' information: http://www.silverstones.com/thebat/TBUDLInfo.html
