On Tue, Mar 01, 2005 at 03:58:55PM -0500, Isaac Richards wrote: > On Monday 28 February 2005 11:30 pm, Gabe Rubin wrote: > Just for the record - For a recommendation engine built into myth, I'd want > the db running on _my_ machine, with the data + server covered under > appropriate oss licenses. 'MythRecommend' doesn't do that.
There are many types of suggestion engines I have been thinking about. Many people are familiar with Tivo's, which takes the log of what you have recorded, or given thumbs up/down to, and remembers things like titles, genres, actors and such. It then scans all new programming and gives weights to how closely they match what it remembers and generates suggestions from the best weights. One could code this for myth without any external data (though right now myth I believe discards much of this information once a show has past, or so I gather from looking at the tables that appear to contain it.) If you want to amalgamate data from other users, I am not sure how to efficiently do that with a server running on your own machine. You could have people just upload anonymized data to an open server which simply gathers the raw data, and then the user's own machine sucks it down and analyses it to generate suggestions. That seems to involve a lot of data flow, or am I missing your goal. Generally I don't think you need to work a lot on finding new series to watch. There are only so many series out there and a lot of existing tools, from reading reviews to talking to friends, help us pick what series we'll check out. An automatic tool might do it a bit faster but also less accurately. Of greater interest are movies and specials and particular episodes. When it comes to movies, there are of course many ways to hunt for those. Once found they can easily be imported into tvwish or even mythtv's own search profiles. (I originally wrote tvwish under another name as just a bulk importer of mythtv search requests, but I decided it made more sense to put these in a standalone program and not attempt a remote-control based interface for them.) Specials are different. Unlike series they are many and change every week, so a system to identify good specials in advance makes sense. The really interesting thing for me is episodes. We know what series are out there, and telling us what people are recording might help us there but what is much more interesting is how _good_ an episode is, which can only be learned after people watch it, or from the newspaper TV critics who get the episodes in advance on tape/dvd. By the way, I think it would be really interesting if somebody became a tvwish style critic, creating a daily file of recommendations to import, and could then go to the studios and say, "5,000 people download my recommendations every day and automatically record what I watch." That might get them on the list to get the advance dvds/tapes just like newspaper critics. Probably help if they did a blog for people who can't auto import the recommendations. (Also on my list is an attempt to add a 'Notes' column to the myth database where people or programs can enter notes about programs. This would include critics adding a note about why they recommend a show, or users of mythtv adding notes for themselves or the other users of their system. Ie. 'Mary, watch this!!!' and so on.)
_______________________________________________ mythtv-users mailing list [email protected] http://mythtv.org/cgi-bin/mailman/listinfo/mythtv-users
