We start with an analytic framework that
has eight dimensions of analysis. And I'll take you through each of these as
we go forward, but this framework and even this slide will be
useful as you will have your own assignment about analyzing a, a
recommender system. We'll be asking you to go through and look
at their the domain. The purpose, the context, and so forth.
So when we talk about domain of recommendation,
what we really want to know is, what's being recommended.
Are we recommending news articles?
Are we recommending products? Are we recommending bundles where instead
of a particular product we're saying look, if you're interested in product x we
can give you a deal if you buy x, y, and z
together. Are we matchmaking, where we're
recommending people to other people. Are we recommending sequences, like a
music playlist, where it may not just be the set of songs, but that the
order of those songs, may matter.
And a particularly intereseting property of a recommendation domain Is the way in
which we treat items that have already been experienced.
In some domains, we're primarily interested in recommending new items,
things you haven't experienced before. If you come looking for a movie, most of
the time, not always, but most of the time, you're looking to see a
movie that you haven't seen.