From mboxrd@z Thu Jan 1 00:00:00 1970 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me X-Spam-Level: X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00 autolearn=ham autolearn_force=no version=3.4.4 X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: 103376,626842fa695f2fdf X-Google-Attributes: gid103376,public X-Google-ArrivalTime: 2004-01-17 06:52:53 PST Path: archiver1.google.com!news2.google.com!fu-berlin.de!uni-berlin.de!hse-mtl-ppp74436.qc.sympatico.CA!not-for-mail From: Christopher Browne Newsgroups: comp.lang.ada Subject: Re: GC, existed? the foreigner Date: 17 Jan 2004 14:52:51 GMT Organization: cbbrowne Computing Inc Message-ID: References: <100glfb4e45iof0@corp.supernews.com> NNTP-Posting-Host: hse-mtl-ppp74436.qc.sympatico.ca (64.229.208.213) X-Trace: news.uni-berlin.de 1074351171 15689932 64.229.208.213 ([125932]) X-Draft-From: ("nntp+wolfe:comp.lang.ada" 5575) X-message-flag: Outlook is rather hackable, isn't it? X-Home-Page: http://www.cbbrowne.com/info/ X-Affero: http://svcs.affero.net/rm.php?r=cbbrowne Xref: archiver1.google.com comp.lang.ada:4492 Date: 2004-01-17T14:52:51+00:00 List-Id: In the last exciting episode, "Randy Brukardt" wrote: > "Christopher Browne" wrote in message > news:bu7i55$ema5s$1@ID-125932.news.uni-berlin.de... >> If none of your _real_ email contains words like "egret," "beseech," >> or "shibboleth," then it certainly won't look like "ham." > > The initial description of Baysian filters included a rule that > anything unrecognized was considered 10% chance of being spam. In > that case, sticking any garbage into a message will help get it > passed. I doubt that current filters work that way, but I don't know > for sure. Which paper provided evidence of the efficacity of that? It seems hard to find an "initial description"; the major papers seemed to emerge in about 1998, and even at that point, they were primarily writing about document _classification_, not "spam detection." I contributed to the work on Ifile back in 1996/1997 (before 1998!), and have been using Naive Bayesian filtering ever since; there is NO such rule in the code I use, and I have never seen such a rule in the scientific literature. Actually, it doesn't even make sense to suggest such a rule. Naive Bayesian filters don't use random number generators to decide what to do with mail; that "rule" can be of _no_ help in what is an entirely deterministic classification process. -- If this was helpful, rate me http://www.ntlug.org/~cbbrowne/ifilter.html "La Cicciolina [...] Electing her was an interesting contrast to the situation in the UK: In Italy they elect a representative from the sex industry. In the UK, they elect their clients." -- Peter Gutmann