; this file contains advanced bayesit settings which may be not interested for all number of users. It must be placed either in the bayesit's working folder, either in the same folder as bayesit.tbp (in this case the new file with copy of settings will be created in the working folder). Values can be either solid (without spaces), either enclosed in quotes. In the case of solid values ';' required at the end of erevy line.
; You can also place the file "bayesit.ini" with basic settings which are available during GUI setup - the settings from the file will be used as default during filter's installation (first run).
; the names of values for bayesit.ini is there:
;working directory (string)
;filter log (string)
;list of ignoring kludges (string)
;show splash screen (1 or 0)
;log current process (1 or 0)
;log global state" (1 or 0)
;log grades (1 or 0)
;filter is logged (1 or 0)
;use white list of kludges (1 or 0)
;use partically translitting (1 or 0)
;sighificant tokens (integer - by default 15)
;spam transliteration table (string)
;local user alphabet (string)
; if 1, export "selective.txt", if 0 - don't export.
export selective download = 1;
; indicates how many times a header token must be met in spam headers to be included
; into "selective.txt" file. Lower value increases filesize and increases possibility
; of false positivies.
selective download spam threshold = 10;
; if token is longer than this value, store it in the base in form of hash
limit size to hash = 19;
; if header token is longer than this value, store it in the base in form of hash.
limit size to hash header = 96;
; trait numeric HTML comments as marks of spam
simple digits spam marks = 1;
; trait any solid HTML comments as marks of spam
no spaces spam marks = 1;
; The number of "!" which will be added subsequently to the token in order to check all it's variations (during degeneration). If previous parameter is 0, this one is ignored.
number of exclamations = 5;
; This value defines the delay (in seconds) which BayesIt! waits before starting autotraining process. Before this value was hard-coded and equal 30sec., but it seems that bigger value (like this 300sec = 5min) is more appropriate. With lesser values BayesIt can begin recalculation of it's dictionary during receiving of a big letter (which needs more than 30 sec. to be downloaded), which could slower mail delivering
training delay = 15;
; if 1, the filter can create it's base "from zero" - i.e. even if no "mark as" action were made for any dictionary, using only incoming mail, by marking only spam "as spam" and leaving non-spam. If 0, the filter can be trained ONLY if at least one letter was marked as spam and non-spam.
learn from zero = 1;
; if yes, the filter will automark all letters according assigned threshold for "move to junk folder" and result of working. If no, you must "mark as" ALL incoming letters in order to keep the filter's base healthy.
use autotrain = 1;
; If this value is 0, the filter works usual way and calculate the final grade of a letter using a bayesian formula. If it is set to 1, the filter just calculates arithmetic average as the final result (which is linear and smooth).
use average instead of bayesian = 0;
; if 1, and the filter can't found token during regarding, it will try some variations of the same token (from capital letter, all caps, same variations from header/body, same variations with added "!"). It helps to improve the quality of filtering. If 0, the filter will use default value for absent tokens. It helps to make the filter faster.
use degeneration = 1;
; This value indicates if the filter will first try to load necessary charset's decoding table from The Bat! registry (the values from HKCU\RIT\The Bat!\XLT) and then (if nothing found) from the file with appropriate name which must contain the same data as the registry, like koi8-r.bin, win-1251.bin, ibm866.bin etc. from the filter's program folder and finally (if it was failed) from the filter's working folder. If this value is zero, the registry checking will be skipped and the filter will use only external files.
use translit tables from registry="0"
; If it is integer number from 1 and more, recalculation is completed after filter collect more than this number of letters. If it is float value below 1 (as 0.001), recalculation is completed after filter collects the quantity of letters which is more than total number of letter which are already in base multiplied to this number (for example, if you have 1000 spams and 2000 hams and this parameter is set to 0.01, recalculation will be when you collect (1000+2000)*0.01 = 30 letters. But anyway in this case recalculation will be done if number of collected letters is more than 100.
recalculating strategy = 50;
; this number shows, how much "heavier" non-spam tokens than spam. It makes some kind of "guard" and keeps from false positives. Usual value is 2, but you can also try others...
regarding threshold = 2;
; This value sets the time interval (in minutes) which is used by server module to automatically shut down, if nothing happens.
;idle time to shut down server="20"
; This value determines what client do when no base exists. 0 - never ask. 1 - ask every time when such situation happens. 2 - interactive (user determines whenever ask or not).
behaviour when no train info="1"
; This value determines the optimal size of every half (spam / not spam) of the base. The training engine will try to keep the number of trained messages no more than this value. During usual work the "new" collected messages - if the total size will preceed this value - will cause the "old" messages to go out of base. If the number of messages is not yet reached this value (i.e., you have only 700 of one sort, and 3000 of another), the engine will calculate the base with smallest of two values (i.e., 700 / 700).
optimal base size per half="5000"
; This value determines the size of "atom" dictinary. The base consists from the atoms with total quantity of messages determined by previous value.
optimal one dict size="500"
; This value determines the minimal size of "atom" dictinary. Usually the filter tries to create the atom dictionaries with "optimal" size. But in the beginning, even the size of one optimal atom dict can cause disbalance. For example, when we have 500 in atom dict of spam, and 10 in no-spam. In such border situation the filter determines the small size and decrease the size of "top" atom dictionary down to this value (but not less). In this case the filter will have choice to set the base in 20 of spam and 10 of non-spam, which is much better than 500/10.
min optimal one dict size="20"
; This value applies when the scanning wizard invoked. If the number of scanned messages related to the size of the whole base is more than this value, then the double scanning will be applied, namely: the messages will not be moved into their determined destination after the scanning; instead the filter will perform autotraining without moving the messages - and only then will actually scan them with moving, using this "just-trained" base.
