Any Neural Net code in Python? I want to filter out spam email

s713221 at student.gu.edu.au s713221 at student.gu.edu.au
Thu Apr 19 05:56:37 EDT 2001


> Personally I feel neural networks are overkill for this sort of thing.
> You're looking for keywords, so I think you're going to need a neuron for
> every possible word. It might work, but I think just making some rules
> manually is easier. You already list a few things you could filter on.
> 
> How many *different* mails are in that bunch of 1869? That's a pretty big
> set, it's at least cool to have a nice training set for things like this :).
> 
> I haven't looked at neural nets for information filtering much, but
> the paper at http://citeseer.nj.nec.com/oard94neural.html (Neural Networks
> in Information Filtering and Retrieval, from 1994) is a list of references
> to papers about the subject that you may find interesting (you can get the
> paper in a number of formats at the top right of the page). There are some
> references to Usenet filtering with neural networks.
> 
> --
> Remco Gerlich

Maybe the best method is to create a user-expandable list of mildly
configurable filters, and use a genetic algorithm or neural net to
decide what filters (and the configeration of each filter) to use. Kind
of mixing an Expert System with a genetic algorithm or neural net. The
thing is, as these tools evolve, the spammers would look for new ways to
trick them, so you're going to need continuous  training. E.g. Spammers
may start sending mail with all lowercase in the subject .... *thinks
for a minute* .... naaaaahhhhhh.

Joal Heagney/AncientHart



More information about the Python-list mailing list