Query failed: connection to localhost:9312 failed (errno=111, msg=Connection refused). Does "Mail" stop learning? - Mac Software Discussions on AppleInsider Forums Toggle navigation All Forums Recent Posts Sign In Does "Mail" stop learning? addison Posted: June 23, 2004 3:58AM in Mac Software edited January 2014 I receive a regular newsletter each month from a supplier. Each month Mail marks it as junk and I click "Not Junk" but Mail just does not learn this.Why? Comments Reply 1 of 6 lucy Posts: 44member June 23, 2004 4:17AM I don't know why the letter keeps being marked junk, but a solution to the problem is to create a custom rule for the newsletter. While it isn't as elegant as Mail learning/not being stupid, it should be fairly easy and effective if the letter always comes from the same address or has the same/a similar title.Hope that's somewhat helpful, even if it doesn't explain your problem with Mail.Edit: To Merovingian: I am fairly certain that I have read somewhere that Mail does not use a Bayesian filter, but rather a more sophisticated/processor intensive filter based on semantics. (I could be wrong.) Reply 2 of 6 merovingian Posts: 436member June 23, 2004 4:19AM The spam filter is a Bayesian filter, which uses mathematical probabilities to determine if an email is spam, by reading the occurences of words and phrases in the message itself. It seems the filter isn't getting enough information to regard it as not junk. In the preferences, (then under "Junk Mail", then click "Advanced..."), you may want to add another rule to say if a message comes from firstname.lastname@example.org, where the address is the newsletter's sender address, then immediately say it is not spam. Hope that helps. m. Reply 3 of 6 stoo Posts: 1,490member June 23, 2004 2:18PM Mail's junk filter isn't as simple a Bayesion network, but is rather more complex. A mail message is a point in the (rather large) wordspace. This dimensionality of this space is then reduced, capturing the general patterns in the messages' words and hopefully getting rid of some noise (which spammers seem to be getting keen on). Cluster analysis is then performed on the data; each each cluster represents a similar topic.http://www.macdevcenter.com/pub/a/ma.../spam_pt2.html(I can't beleive I studied things like latent semantic analysis and singular value decomposition less then two years ago... ) Reply 4 of 6 scott Posts: 7,431member June 23, 2004 7:36PM You don't use SVD in your line of work? Reply 5 of 6 arnel Posts: 103member June 23, 2004 11:46PM Do you have "Trust Junk Mail headers set by your ISP" turned on? I'm wondering whether your ISP is flagging the message as junk, and therefore Mail is ignoring any of its own rules and just trusting that...Neil.a.k.a. Arnel Reply 6 of 6 addison Posts: 1,185member June 24, 2004 2:28AM I am aware that I can create a rule but I don't understand the purpose of the "not Junk" button if the application ignores it.Yes I do have trust my ISP switched on, but as I run my own mailserver mx records et all I don't think my ISP touches the incomming mail at all. Sign In or Register to comment.