OK, I’ve been living with this for too long, and I can’t find anything about it anywhere with web searches.
Thunderbird’s spam filter is absolutely stupid. Every effing day I have to dig through the junk folder to find legitimate email from people whom I’ve repeatedly (as in dozens of times) told it are not spammers, and from subjects (e.g., mailing lists, like arocket) that I’ve repeatedly told it is legitimate email I want to read. It absolutely refuses to learn. I can handle false negatives in spam reporting, but when it continually buries legitimate emails after I’ve implicitly whitelisted them, I can no longer tolerate this.
Am I the only one with this problem?
According to their knowledge base…Bayesian filtering requires at least 100 bad messages be marked as spam AND 100 good messages marked as not junk to function. To work best, it needs a few hundred of each marked. Just checking to see if you have also marked at least a hundred good messages as such.
I’ve never seen a Bayesian filter that requires that much training. I used to use an Outlook add-in that worked about 98% out of the box.
I use the Eudora front end for Thunderbird and haven’t noticed any problems.
I’ve dumped Thunderbird once and for all. One of the reasons was that I ended up turning off the spam controls altogether. It made no sense to keep moving false negatives to the junk folder to train the damn thing when it’s so much simpler to delete them when I find them — which defeats the purpose.
Now I run all of my email through Gmail which already has a well-trained filter. I can set up my phone to use Gmail without using the Gmail setup so that I can set my outgoing address as non-Gmail, and if I want to click a mailto link in Firefox I’ve installed an extension that opens a Gmail compose window instead of trying to open a local mail client.
Rand:
Nope — I’ve had the same problem for some time and had even reached the point of having my webserver forward all my mail to my Gmail account in order to use their spam filter.
My eventual solution, though, was to pay for SpamFighter (I think I paid $59 for a 3-year subscription). Once installed and running — and after just a little additional training — it started filtering out probably 95%+ of the spam that TBird was letting through. It made TBird usable again. ..bruce..
My concern is not the false negatives (spam getting through) but the false positives — emails I need to see that I have to repeatedly dig through the junk for. Not whitelisting someone whom I’ve told it is not a spammer is inexcusable.
Why not just go with gmail? their spam filter is superior…I have been a gmail user since very early on, works like a champ.
Because I want to a maintain control over my own mail server, and don’t trust Google?
What are you running for an email server? Can you have the server do the spam checking and set Thunderbird to “Trust junk mail headers set by: SpamAssassin/something else”. Then disable adaptive junk mail controls for the account.
I’m running SpaceAssassin. I didn’t know that TBird had that option.
I get more false positives from gmail than thunderbird. A massive training set of emails might make the difference. I have the email provider do primary spam filtering, then set thunderbird to just flag spam, not move it to the spam folder.
I’ve had false positives before, but had no problem after telling Tbird they weren’t spam.
Something I’ve done in the past when a Baeysian spam filter goes awry is to dump the training set and start over. Performance will be worse at first, but then improves as you populate it with fresh inputs. YMMV.