This is the point where great content recommendation comes into play. Read the information about the NetFlix prize that has been on the front page or just go over to http://netflixprize.com for more info.
This is one of the aspects I've been looking into using Content Recommendation for, but I'm looking at it from a larger information overload perspective.
Even a site like Digg can only really do popularity because of the binary nature of the thumbs up/thumbs down system, but once you move past the binary scale into a 5 star rating system and take your algorithms from a slope one into a latent semantic indexing, or something similar, instead of popularity you end up with personalization.
Of course this all depends on the quality of the data and the size of you user base, plus the cross section of the users that have voted on similar items, in this case blogs.
I'm personally taking the bet that this approach is one of the next great ways to organize information and is worth the next few years of my life to work on .
I am trying to solve the problem of blog consumption. Blog consumption, by nature is personalized. You have your own set of feeds that you read daily. I have a different set. Which means that whatever feed/post appeals you, might not suit my taste, and vice versa. Thus, community rating like digg etc, is not the right way to treat your blog posts, or in other words, you should not draw your recommendations "completely" from such a system. That's why i intend to build a personalized system, (may be plugged in your favourite reader) for this!.
Current blog discovery mechanisms like technorati and google search are not much helpful in that. Consuming blog feeds has more to do with personalization. Keyword based search is a bit callow for this. ideas use case should be when u r reading a blog entry, your reader should throw recommendations for it. I am planning to build this infrastructure. Any suggestions/critics for this?
You could implement the recommendation backend so that it can be accessed from feed readers, server-side frameworks, client-side widgets, and browser extensions. This way you can also gather data from all these sources to base the recommendations on. Perhaps there's a revenue-sharing scheme to motivate third parties.
I guess finding similar blogs would take more long-term data whereas finding similar posts should react very quickly. I'm specifically thinking of the case where two blog posts are both linked to by a third post: those two could then be considered similar. Comparing the frequency of some key words mentioned in the posts would be another way to get started, as would be looking at the audiences.
Whatever the method, it should provide results that users find meaningful in meeting the need they have.
This is one of the aspects I've been looking into using Content Recommendation for, but I'm looking at it from a larger information overload perspective.
Even a site like Digg can only really do popularity because of the binary nature of the thumbs up/thumbs down system, but once you move past the binary scale into a 5 star rating system and take your algorithms from a slope one into a latent semantic indexing, or something similar, instead of popularity you end up with personalization.
Of course this all depends on the quality of the data and the size of you user base, plus the cross section of the users that have voted on similar items, in this case blogs.
I'm personally taking the bet that this approach is one of the next great ways to organize information and is worth the next few years of my life to work on .