ok, so in the future Im going to need to setup a bayesian filtering system on a website.
Its going to need to provide different filtering based on which user is accessing it.

now, currently I dont have much experience with bayesian filtering, so partly Im just starting to figure out the right questions to ask...on that basis any response at all will prolly be useful.

So, assuming that the bayesian filter stores everything in a database at the back end, would I need to literally create a new database for each user?  to copy the files?  \

how would you do this?  I need a separately trained filter for each user, all available over the web.

each and every user?

not user clasifications?

You could have a field called "user"

But that's one of those newfangled relational thingos...

Philo

"not user clasifications?"

huh?  how does that work?

"You could have a field called "user" "

how does that work with the bayesian filter though?  sure I can setup a database with the username....but how do I ensure that each separate user gets a clean bayesian filter of their very own to train?

(actually probably they will need multiple bayesian filters each of their very own)

I was thinking if you knew how many different 'types' of users you were going to have, you could just pass the entries through each filter 'type' and store those scores.

ah, I see.  no, not types. we will need separate and clean bayesian filter(s) for each and every user.

Look at Reverend its a standalone Bayesian filter.  If you use the user id to create a corpus then just use that user id to specify which corpus to use when filtering.

wow, that looks perfect thanks Simon.  its even in python :)  Ill download it now and take a look.

It even works.

That't the basis of what I'll use over there.