jbayer at gmail
Apr 4, 2006, 2:48 PM
Post #1 of 1
(542 views)
Permalink
|
Hi,
I'm interested in adding user ratings for programs. First of all, is anyone
currently working on this or has it been considered but nixed in the past?
A summary of a good ratings system (TiVo) is at
http://scholar.google.com/url?sa=U&q=http://portal.acm.org/ft_gateway.cfm%3Fid%3D1014097%26type%3Dpdf
In a nutshell, they presented:
- explicit thumbs [up/down] for a series ... which implicitly leads to
thumbs for actors/genres within the show
- implicit thumbs based on recording a series or watching livetv, but this
can be extended to other kinds of behavior as well
- recorded suggestions based on:
+ shows with positive thumbs but are not currently being recorded
+ content of the thumbs (ie you might like similar genres actors and so
forth)
+ collaborative algorithms on a central server or decentralized system
I would suggest developing this in the following order:
1) schema + insertion algorithms for "thumbs." Support explicit and implicit
thumbs, and for series, genre, actor (+ other types)
2) allow users to rate programs while playing. This should change the thumb
count for that series AND any genres/actors associated with that program.
3) optionally mark implicit ratings, such as recording a show (see TiVo
paper)
4) suggestions using [explicitly rated but not recording] and
[content-based]
5) optionally, and with opt-in, have a collaborative filtering mechanism.
Question about if it should be centralized [and who would run this] or
distributed.
I would be more than happy to work on this or work with someone else on
this. I for one would certainly find this useful. Any thoughts or
suggestions?
Apologies if this idea already has been discussed. I found one thread in
Sept 2003 titled "'recommendations' section" and Charles Brej has been
working on letting users explicitly recommend shows via mythWeb (but it
lacks some of the infrastructure for weightings, types of recommendations
(series vs actors), implicit recommendations, etc. If I understand his
changes correctly it can still be quite useful and reused for recommendation
frontend changes.
Also, what is the development process? Ideally, I feel it would be nice to
get smaller chunks in in the order I listed above, as opposed to one monster
change. The makes the changes manageable, but the downside is that it's
possible some of the pieces are checked in and done but the rest don't
happen.
Thanks,
Jason
|
