• My kSpam for Linux ignores the KL_BL_IGNORE setting, causing false rejections

    By Jörg Asmussen 2 decades ago

    I have put the folling line in NOTES:INI, but I still get many hits on tokens mentioned below:



    KS_BL_IGNORE=10px,13px,arial,background-color,band,bgcolor,bold,boulder,cellpadding,cellspacing,colspan,div,FF0000,ffcc00,FFFFFF,font-family,font-size,Helvetica,helvetica,icons,meta,nowrap,rowspan,sans-serif,text-decoration,url,verdana,Verdana,



    How can I verify if kSpam is acknowldeging the settings?



    Best regards

    Jörg

    • I think you must add a space

      By Hans-Georg Franke 2 decades ago

      after the colons. (Don't use this by myself, but i will try it)



      Greetings Hans-Georg

      • Do you mean like this ... ?

        By Jörg Asmussen 2 decades ago

        KS_BL_IGNORE=10px, 13px, arial, background-color, band, bgcolor, bold, boulder, cellpadding, cellspacing, colspan, div, FF0000, ffcc00, FFFFFF, font-family, font-size, Helvetica, helvetica, icons, meta, nowrap, rowspan, sans-serif, text-decoration, url, verdana, Verdana,



        i.e a space after each comma?

        • You...

          By Tom Lyne 2 decades ago

          will not see them in the KS_BL_TOKENS item if it's working.



          There is no space after the comma.



          -tom

          • I have a line as specified, without spaces ...

            By Jörg Asmussen 2 decades ago

            … but the words do still appear in the KS_BL_TOKENS item. Is there a way to verify which tokens are ignored once the BL is running?

            • The tokens might still appear

              By Tom Lyne 2 decades ago

              if they are the only words in the email.



              Bload ignores the tokens when it calculates the probabilities, but if an email has just those tokens then they will appear in KS_BL_TOKENS but with a probability of 0.5.



              -tom

              • By Jörg Asmussen 2 decades ago

                Is it case sensitive?



                This is a list of current hits (Tokens-value=0,99)



                0001pt

                0in

                0pt

                25in

                3DContent-Type

                3DGENERATOR

                3Dtext brother

                cellPadding

                cellSpacing

                colSpan

                Definitions

                distributors

                felony

                FONT-FAMILY

                FONT-SIZE

                Generator

                helvetica

                margin-bottom

                MsoNormal

                purple

                SPAN

                TABLE

                TBODY

                TEXT-DECORATION

                underline

                vAlign

                Verdana



                here is my

                KS_BL_IGNORE=10px,13px,arial,background-color,band,bgcolor,bold,boulder,cellpadding,cellspacing,colspan,div,FF0000,ffcc00,FFFFFF,font-family,font-size,Helvetica,icons,meta,nowrap,rowspan,sans-serif,text-decoration,url,verdana,





                Btw, I have created a new view in MailGood where I show false negatives (moved from mailspam to mailgood) and show the triggering tokens in a categorised column with these design changes:





                Select formula:



                list0 := KS_BL_TOKENS;

                list1 := @ReplaceSubstring( list0; ": 0,9900"; "");

                list := @Trim(@Replace(list1; list0; ""));



                SELECT list != "" & @ToNumber(KS_BL_PROB) > 0,5 & (!@Contains(confirmation; "O"))



                Categorised column



                list0 := KS_BL_TOKENS;

                list1 := @ReplaceSubstring( list0; ": 0,9900"; "");

                list := @Replace(list1; list0; "");

                @Trim(                list)

                • Yes, it's case sensitive.

                  By Tom Lyne 2 decades ago

                  The Bayesian filter is the only part of kSpam that is case sensitive.



                  -tom

                  • Is this intentionally?

                    By Jörg Asmussen 2 decades ago

                    How many words can be added to the KS_BL_IGNORE line?

                    Who codes the BL? wouldn't a simple @Lowercase (or whatever the equivalent function in C might be) enable caseINsensitiveness? Maybe adding a Variable KS_BL_NoCase=1?