I typically will look at source view of an email if I can’t determine if it is spam and if Spam Siever does not catch it.
In the last couple weeks I’m getting lots of spams that SpamSieve is not catching and when I look at the source view, it contains ASCII art messages such as “Beautiful Babes”. The subject of the message is “SEXUALLY-EXPLICIT: killer kansas birdie hospital hand certainly RN8DtXyfkpIvD6a53kfzXA” but SpamSieve is not catching these and I’ve tried training numerous times.
The next step is to check the log. For what it’s worth, it sounds like these messages are obviously spam, so SpamSieve probably isn’t missing them; there’s likely a setup problem.
Right after I made the forum post, I sent an email along with the log file and the source of the email to <spamsieve-fn@c-command.com>.
As I looked through the log, it looked like SpamSieve was catching the exact same types of Subjects most of the time, yet for reasons I don’t understand, sometimes they slip through. I trust you’ll be able to figure this out Michael. thanks.
The log shows that SpamSieve wasn’t asked to filter the ones that got through. You can see this because the log shows that they were “Trained: Spam (Manual)” but not “Predicted: Good” or “Predicted: Spam.” Therefore, you have a setup problem in Entourage. Perhaps the SpamSieve rule(s) aren’t setup properly, or maybe you have other Entourage rules above them, or a Mailing List Manager rule that’s interfering.
I have reinstalled the scripts, I purged the corpus and retrained with 400 spams and 350 good email.
I have to the best of my knowledge, followed to the letter all of the steps to make things right.
Now I’m seeing messages that are in my Junk folder that the log says are “Predicted-Good”. How are these ending up in my Junk Folder? What would cause this? I do have some other mail actions running after the first two SpamSieve actions, but non are configured to move anything to Junk folder. I have also made sure that Entourage 2004’s Junk email filer is off. and I do not have any mailing list rules.
UPDATE:
My bad. I just discovered that I had overlooked changing the second spam sieve Move Messages rule cagtegory to “Junk” from “None”. This, of course, was moving all good messages to the Junk Folder.
Just to be clear, you should only reset the corpus if the accuracy is poor and you have verified from looking at the log that SpamSieve is predicting that the spam messages are good. This was not the case with your log.
Does this mean that everything is working correctly now?
My original corpus was fairly new anyway since I had only recently installed spamsieve on a fresh installation of Leopard. I never trained SpamSieve when I installed it on my new system so actually I’m glad to have started over today with it.
As of right now, things seems to be OK. Predicted Good messages are no longer going to the junk folder and in the short time since fixing my mistake, all spam that has come in so far has been correctly identified and moved to junk folder.
I’ll keep an eye on things and report back if anything is awry.
Thanks Michael for such a helpful app! It’s the best…only the operators of the software need to improve
As a follow-up I’d like to report that SpamSieve is operating perfectly now. It’s been extremely accurate and the problems I had reported were due to my errors setting up SpamSieve.
