Search  Rules  Login
TheBat English » The Bat! - Filters
Pages: 1
RSS
Problem with Bayesit
 
#1
0  
I use Bayesit for filtering spams. I recieve some 150 spams a day. It filter quite all the mails as spams after short learning. Even the good ones: false positives. Marking mails as spam is working ok, marking a mail as not spam does nothing: the next mail from the same sender goes to Junk as before.
So, for now, junk folder just replace the incoming one ;-)
Is there is a way to get this strange anti spam working ?

Is Bayes Filter plugin better ?
Can i use the both together ?

Thanks in advance.
 
 
 
#2
0  
I don't dare to say which is better, but you can use more of them.
You can set TB to pick the average score of the used plugins, the highest score or the lowest score to act upon. You can set that at:
Options -> Preferences -> Anti-spam
__________________________________
I'm just a user of The Bat! I don't work for Ritlabs.
 
 
 
#3
0  
Thanks. I used K9, witch is a bayesian proxy, and was working perfect. It is a pity that The Bat! don't have a good working antispam.

BTW, when you click on "mark as not spam" on a message in Junk, the bayesit would better moove the message in the incoming folder, as it moove to junk a message you mark as spam, from the incoming folder.
 
 
 
#4
0  
What made you decide to quit K9 then?

I can understand why you'd say that marking as 'not spam' should be acompanied by an action to move the message to the inbox, but that would definitely mess up things in the training stage.
When you're still training the plug-in you're supposed to feed it a load of good and bad messages. Most of my messages get filtered into other folders, so if marking them as good would cause them to be diverted back to the inbox I'd quit training (and therefore using) the plug-in rather soon.
Same goes for false spam in the junk folder, I mark it as good and before I can place it where I'd like it, I'd need to find it again in the Inbox, that would be a major disadvantage too.
And finally when you're using a common junk folder, the association with the originating account gets lost, so it wouldn't be possible to drop it into the right Inbox.
That's three reasons why you wouldn't want the 'mark as not spam' action to move your message to the Inbox.
__________________________________
I'm just a user of The Bat! I don't work for Ritlabs.
 
 
 
#5
0  
Quote
Roelof Otten wrote:
What made you decide to quit K9 then?
Because it is a proxy: an other app, listening *allways* on the port (even when your client is closed), and that you have to open in order to configure, correct false positives, or mark  mails as good/bad.
Not so simple as inside the mail client.

Anyway, Bayes-it is not working at all. An exemple: I recieved a mail from a friend in raw text with no spams words inside. Nothing like a spam, and in french. This mail was sended in junk !
So, i marked-it as good, moved it in the incoming folder and answered-it (and i add the adress of the sender in adress book). When i recieved the next answer from the same adress, it was again in junk, again and again.
False positive is the worse an antipâm can do. And i wonder why, when you mark a spam as good, the sender adress is not added in the white list.
What the hell is doing this "mark as good" thing ?
 
 
 
#6
0  
You need to train it by feeding it enough good and bad messages, it adds the necessary words to your spam/ham dictionaries. I believe it's possible to whitelist addresses, but I don't know how, because I don't use the spam plug-ins anymore.
__________________________________
I'm just a user of The Bat! I don't work for Ritlabs.
 
 
 
#7
0  
Quote
Roelof Otten wrote:
You need to train it by feeding it enough good and bad messages
When you mark a mail as good, even only one time, other mails from the same sender would Never be lost. NEVER !
Ok, i can conclude that the Bat! (or voyager) has no usable antispam system, and i understand why you don't use antispam plugin anymore ;-)  
 
 
 
#8
0  
I've never used any plug-ins with Voyager as most plug-ins store stuff in the registry, that's not a thing when you're playing around on multiple computers (not necessarily your own).
__________________________________
I'm just a user of The Bat! I don't work for Ritlabs.
 
 
 
#9
0  
Esperado is right. I am having the exact same problem.

No need to explain.

BayesIT is a crap period.

Now that we know that, what's the solution?
 
 
 
Pages: 1