Re: Bayesian SPAM filtering
Hello James Olsen, On Sun, 23 Feb 2003 09:10:02 -0600 GMT your local time, which was Sunday, February 23, 2003, 10:10:02 PM (GMT+0700) my local time, James Olsen wrote: Hello Stefan, ST The thing stated in the subject can be supported if we provide some ST additional features for plug-ins like adding menu commands, am I right? That would make supporting it easier. I'm not sure how much you know about Bayesian filtering, so forgive me if this sounds remedial. Bayesian filters need to be fed spam and good email messages and it begins to learn the difference between the two. seems to me we have to start again collecting mail oops spam... -- Best regards, tracer Using The Bat! v1.63 Beta/7 on Windows XP 5.1 Build 2600 Service Pack 1 Current version is 1.62 | Using TBDEV information: http://www.silverstones.com/thebat/TBUDLInfo.html
Re: Bayesian SPAM filtering
Estimados seguidores del tbdev arroba thebat.dutaint.com: En relación a lo que Stefan en su momento posteó: ST The thing stated in the subject can be supported if we provide ST some additional features for plug-ins like adding menu commands, ST am I right? we are replacing the menu commands with a external configuration utilities, today i need more information of the mail and the user from the bat!: - the size of the mail [KB] - the decoded body and the list of atachments. (separed with ',' please) - a list of mails from the address book's. in www.gdutb.org we was talking about a two times spam filter: ¿what is this? if we have a second method to know the score like TBP_GetScoreBeforeDownloading, we can filter the mail with the sender, the subject, and the kludges, the list of attachments, say to the bat! the score and with this you decide if the mail was donwloaded or deleted in the server whitout downloading. (like selective downloading) Next with the actual interface, the plug'in analizes the complete mail to define a score, and now it's works like 1.63beta3 to beta7. -- Se despide, Task Control mail: TaskControl at SoftHome dot net correo: TaskControl arroba SoftHome punto net Usando: - Windows 98 4.10.1998 - AVG 6.0 Free Edition - The Bat! 1.63 Beta/7 - Trillian PRO 1.0 B Current version is 1.62 | Using TBDEV information: http://www.silverstones.com/thebat/TBUDLInfo.html
Re: Bayesian SPAM filtering
Hello Stefan, ST The thing stated in the subject can be supported if we provide some ST additional features for plug-ins like adding menu commands, am I right? That would make supporting it easier. I'm not sure how much you know about Bayesian filtering, so forgive me if this sounds remedial. Bayesian filters need to be fed spam and good email messages and it begins to learn the difference between the two. So, a user needs a way to submit large quantities of both. This could be implemented a number of ways.. Perhaps two new Delete functions: Delete as SPAM or Delete as HAM (a good email). For example, I'm now taking my spam folder, saving it to mbox format, copying the mbox to my linux box, and then running the learn spam function. Then I take my good email from my inbox and follow the the same process, except I use the learn ham function, of course :) I also have to do this for each TB folder (where filtered mail goes for mailing lists, etc) since I can't export multiple folders into a single good email MBOX file. Ideally (and generically, not-Bayesian specific), plugins would be able to operate on a single message or a list of selected messages. They'd also need some kind of API to to tell TB what to do. Delete the message, copy the message, mark it in some way. I'd also like to see support for DLL plugins, not just EXE. I realize that might make implementation a bit tougher, but when you start calling a filter or plugin on thousands of messages, a DLL will take a fraction of the time it's EXE counterpart to operate on all of them since the computer doesn't need to set up a subprocess, run the program, and then clean up and free up memory, for each message. Anyhow, whatever you guys decide I'm sure will be a great help and I appreciate that you are listening and asking for opinions. Thank you! -- James Current version is 1.62 | Using TBDEV information: http://www.silverstones.com/thebat/TBUDLInfo.html
Re[2]: Bayesian SPAM filtering
Estimados seguidores del tbdev arroba thebat.dutaint.com: En relación a lo que James en su momento posteó: JO Bayesian filters need to be fed spam and good email messages and JO it begins to learn the difference between the two. So, a user JO needs a way to submit large quantities of both. This could be JO implemented a number of ways.. [...] learn ham function, of JO course :) [...] The easy way: Today we can select a lot of messages and we have a secondary mouse button, with a lot of options... add two: it is spam and it is a good mail, when a user select a mail (or a group of mails) and click (or hot-key) the options you export this to a folder in a simple text format (whitout attachments please). When a plug'in starts, it can analize the folder, learn about the patterns, add it to anti-spam plugin database and clean the folder. (or two folders) Maybe you can make a similar function to capture texts from the spam, the user mark the word (or phrase) secondary-mouse-button, click in two options: it's a bad string, it's good string, and export this texts to a simple text file. - for vampire if you put a third option: it's a suspect string my work (and the work of vampire users) will be easier - Thanks a lot. -- Se despide, Task Control mail: TaskControl at SoftHome dot net correo: TaskControl arroba SoftHome punto net Usando: - Windows 98 4.10.1998 - AVG 6.0 Free Edition - The Bat! 1.63 Beta/7 - Trillian PRO 1.0 B Current version is 1.62 | Using TBDEV information: http://www.silverstones.com/thebat/TBUDLInfo.html
