Re: Adaptive scoring with "Plan for spam" in mind

Re: Adaptive scoring with "Plan for spam" in mind

[Johan Bockgård]

Dmitry Astapov <adept@umc.com.ua> writes:

> I wonder: would it be possible to modify word-based adaptive scoring
> the way spam.el/ifile/spamoracle works? Right now score for separate
> words are inc- or decremented a fixed amount, but I think that it is
> possible to feed them (words) to Bayesian filter, which then
> modifies (replaces) scoring rules according to his findings.
>
> Does my idea sounds right or I've been reading spam.el too long? :)

Yes (the former).

(info "(gnus)Adaptive Scoring")

        Note that the adaptive word scoring thing is highly
        experimental and is likely to change in the future. Initial
        impressions seem to indicate that it's totally useless as it
        stands. Some more work (involving more rigorous statistical
        methods) will have to be done to make this useful.

-- 
Join us on #gnus @ irc.freenode.net.
http://www.emacswiki.org/cgi-bin/wiki.pl/InternetRelayChat

Re: Adaptive scoring with "Plan for spam" in mind

[Ted Zlatanov]

On Tue, 05 Aug 2003, adept@umc.com.ua wrote:
> 
> I wonder: would it be possible to modify word-based adaptive scoring
> the way spam.el/ifile/spamoracle works? Right now score for separate
> words are inc- or decremented a fixed amount, but I think that it is
> possible to feed them (words) to Bayesian filter, which then
> modifies (replaces) scoring rules according to his findings.

I don't think isolated words will be useful.  Statistical scoring
works best on whole articles.  In fact, I've been using word-based
adaptive scoring for a while without any visible benefit, so I'm
considering turning it off - the underlying concept of adaptive
scoring may be broken, not just the words vs. article issue.

You can consider *all* low-score articles spam (by modifying the list
of spam marks to include the low-score mark, not just the spam-mark,
in the groups or topics where you want this behavior).  That would
work nicely with regular and adaptive score files, since you'll be
effectively telling the spam processor about things you don't like to
read for various reasons.

Ted