On Wed, 21 Nov 2007, Christian Hauschke wrote:
> I'm searching for a FLOSS recommender system for OPACs in general and
> PICA/OCLCs OPC4 in special. What I have in mind is as follows:
> - it should collect data of the titles that were looked at in one
> research session (however defined)
> - there should be some kind of statistical analysis of all the data
> Does something like this exist? Is there maybe something that could be
> used as a base for such a script? I'm new to the topic, but it doesn't
> seem like rocket science to me.
So far as I know, the kind of statistical analysis you're describing is
usually done in some sort of scheduled offline processing task, since it
typically has to "have in mind" a whole dataset at once, so to speak.
Point being that you could implement such a system, then provide a
for this one title, knowing what I know about this user), but I don't
imagine you could actually do all of that analysis in a single run-time
transaction without usability (responsiveness) suffering.
That said, if you're interested in building a recommender system, there's
a new oreilly book on the topic:
I've seen pieces of free software toolkits for this stuff over the years,
but since getting this stuff right tends to depend a lot on the shape of
your own data, there might not be many reliably good general-purpose tools
that would fit this task well.
Hopefully I'm wrong on that part, though!