![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Recommendations for everything
Nov. 10th, 2011 11:29 amThe biggest problem with enjoying good forms of art is finding them. I'm certain that the world is already full of top-quality works that would let me fill my entire spare time with nothing but nonstop moments of awesomeness, if I only knew what they were.
This is, of course, a well-known problem. And there are various recommendation systems that try to figure out your tastes and suggest other things that you might enjoy. Last.fm provides suggestions about music, Netflix has its famous recommendation system for movies, Amazon recommends books plus other products the site sells, et cetera. But there is - at least as far as I know - no recommendation system for e.g. fan fiction.
Having several different recommendation systems seems wasteful. On the other hand, yes, if you build a system for only music, then it's easier to tune it exactly to the domain of musical recommendations, using a variety of rules that only make sense for music. But on the other hand, it seems like at least some of your preferences in one domain ought to be correlated to your preferences in another domain. If you listen to lots of sappy romance songs, then perhaps you'd also like to read some romance novels. Or if you buy a lot of fantasy RPG books, maybe you'd like some fantasy art. With more information available, it would be easier to identify your tastes and anything that made you different from the median romance or fantasy fan.
So I would like to have a recommendation system for everything: books, music, movies, computer games, scholarly papers, fan fiction, amateur fantasy CGI art, adult pictures, IRC channels, interesting people or hobby groups in your neighborhood, you name it. I could just specify a category, and the system would throw out a number of suggestions.
This would, of course, require an immense amount of information to produce anything worthwhile. But that information is already being collected, it is just not available in one place. Google and Facebook already have immense amounts of information about what the tastes and preferred websites of many people. One of them could partner with e.g. Amazon, Last.fm and other similar sites to combine the user information to one big recommendation engine.
Alternatively, many of the existing recommendation sites already offer APIs or other ways for third-party sites to pull the existing information. If they don't already offer it, they might be persuaded to do so. This opens up the possibility for a start-up built around the "recommend anything" concept, whose users could just specify the sites they wanted their information to be taken from.
Obviously there are various privacy considerations to take into account. At a minimum, no information should be shared between two separate services without the explicit permission of the user. Amazon shouldn't be allowed to turn over your shopping history to Google unless you specifically asked it to do so.
Still, if such hurdles could be overcome, and the recommendation system could be made to work well, it could become absolutely awesome. No more wondering of "I'm bored, what would entertain me?" ever again.
This is, of course, a well-known problem. And there are various recommendation systems that try to figure out your tastes and suggest other things that you might enjoy. Last.fm provides suggestions about music, Netflix has its famous recommendation system for movies, Amazon recommends books plus other products the site sells, et cetera. But there is - at least as far as I know - no recommendation system for e.g. fan fiction.
Having several different recommendation systems seems wasteful. On the other hand, yes, if you build a system for only music, then it's easier to tune it exactly to the domain of musical recommendations, using a variety of rules that only make sense for music. But on the other hand, it seems like at least some of your preferences in one domain ought to be correlated to your preferences in another domain. If you listen to lots of sappy romance songs, then perhaps you'd also like to read some romance novels. Or if you buy a lot of fantasy RPG books, maybe you'd like some fantasy art. With more information available, it would be easier to identify your tastes and anything that made you different from the median romance or fantasy fan.
So I would like to have a recommendation system for everything: books, music, movies, computer games, scholarly papers, fan fiction, amateur fantasy CGI art, adult pictures, IRC channels, interesting people or hobby groups in your neighborhood, you name it. I could just specify a category, and the system would throw out a number of suggestions.
This would, of course, require an immense amount of information to produce anything worthwhile. But that information is already being collected, it is just not available in one place. Google and Facebook already have immense amounts of information about what the tastes and preferred websites of many people. One of them could partner with e.g. Amazon, Last.fm and other similar sites to combine the user information to one big recommendation engine.
Alternatively, many of the existing recommendation sites already offer APIs or other ways for third-party sites to pull the existing information. If they don't already offer it, they might be persuaded to do so. This opens up the possibility for a start-up built around the "recommend anything" concept, whose users could just specify the sites they wanted their information to be taken from.
Obviously there are various privacy considerations to take into account. At a minimum, no information should be shared between two separate services without the explicit permission of the user. Amazon shouldn't be allowed to turn over your shopping history to Google unless you specifically asked it to do so.
Still, if such hurdles could be overcome, and the recommendation system could be made to work well, it could become absolutely awesome. No more wondering of "I'm bored, what would entertain me?" ever again.
