ANN Advice

Hello,

I am interested in developing an ANN that will help to predict documents
that staff may be interested in downloading from our large intranet library
system.

At the moment I can gather data from the users of the system such as: the
document summaries they browse to, the document bookmarks they decide to
make, and the actual documents they download to read on their PC.

I would like to be able to begin predicting new documents that users may
want to bookmark or download, based upon their current browsing
pattern/history and the general browsing pattern/history of all other users.

Is there a particular ANN type that would be suited to such a task? Can the
ANN learn as it goes instead of training the system up front? Does anyone
know any examples or tutorials that deal with solutions to this kind of
problem?

Thank you,

Jamie Burns.