[mythtv] Re: Myth recommendation system

[Charlie Brej]

Just keeping the interested people up to date on the current state of the
recommendation system I am working on and explain the goals of the system.

Goals:
The idea is to have a full Tivo Like system which makes recommendations and even
records stuff for you on a whim that you might like. To do this the users must
rate shows and submit this info to the central server. They are then grouped
with other people with the same taste and the other peoples recommendations are
downloaded. People make recommendations both before the show is on and after.
Before the show is on these are used to guide other people to shows. After the
show the reviews guide people as to whether they should bother to watch the show
which was recorded. Even drop its auto expire rating if it was found to be poor 
after recording automatically.

Current state:
The recommendations can be fetched in an xml format and added to the user's
database. These can be browsed though list in lists and appear in the guide
grid. They can also be used to record automatically anything which is
recommended (by a particular source) using the advanced search.
To make recommendations there is an interface in mythweb.
e.g. http://brej.org/mythtv/recommendations/program_detail.php.htm
Exporting and importing of the recommendation set can be done through
mythfilldatabase. This means for now you can publish your recommendations and
exchange them with ones from your friends. See what they thought of last night's
show and what they think you should watch. To try this out I also wrote a couple 
grabbers of TV Picks from websites for the UK and the US.

ToDo:
- Reading and writing the reviews/recommendations in MythTV. (currently only in
mythweb)
- Method of rating recommendation sources.
- Centralized server repository for exchanging recommendations (http based at 
first, then later also a binary direct communication).
- People grouping/recommendation generation server code.
- An automated exchange of recommendations (within MythTV rather than via xml 
files).
- Training of the server (some AI involved)

If I am way off the mark then please say so.

-- 
         Charlie Brej
APT Group, Dept. of Computer Science, University of Manchester
Web: http://brej.org/                 Tel: +44 (0)161 275 6177
Mail: IT303, University of Manchester, Manchester, M13 9PL, UK