On Wed, May 22, 2013 at 1:16 PM, Gabriel Gunderson <[email protected]> wrote:
> On Wed, May 22, 2013 at 12:59 PM, Michael Torrie <[email protected]> wrote:
>> I never could approach the level of Google's Gmail spam filter.
>
> This is great feedback. Thanks.
>
> I do recall being relieved when I didn't have to manage SPAM filtering
> any more. Funny thing is, the more challenging it sounds, the more I
> want to take it on... I think we're getting back to Corey's Q about
> having any sense :)
Well, it is essentially a machine learning problem. Modern machine
learning techniques are very effective at detecting patterns and
classifying based on them, but such techniques are only as good as the
data they are built upon. Google has a pretty big advantage here, as
they've both got industry experts at machine learning and a gigantic
corpus with which to train and tune their system.
If you are generally interested in machine learning rather than just
trying to stop spam, it might be an interesting exercise to perform
unsupervised machine learning on collections of email to see what
kinds of clusters turn up for different ways of doing feature
selection from an email. You could turn up some interesting and
useful new email management tools beyond simple spam reduction.
--Levi
/*
PLUG: http://plug.org, #utah on irc.freenode.net
Unsubscribe: http://plug.org/mailman/options/plug
Don't fear the penguin.
*/
