Which Italian Bistros can you recommend?

Which type of recommendation engine should i choose if i want a cross content recommendation, which means i can recommend you which music to listen to based on which movie you've watched or which news you've read?

• I am working on a multimedia website where you can read news, stream movies , buy ebooks and listen to music.

• Answer:

  I would suggest viewing this as one big recommendation system that includes music, movies, books, etc. Then your problem just amounts to filtering results; when you query for recommended movies, just don't consider items that aren't movies. If you are building on Mahout, or something like Myrrix, you can use the Rescorer interface to implement run-time filtering. The only question you face is how to weight data from different sources differently. It may be that you have rating data for all of these things and they're all roughly comparable. But maybe you have clicks and views and downloads. A news article click is not the same as a movie view. This comes up even within one domain: watching all of a movie is different than watching the first 5 minutes. For that reason I suggest looking at an algorithm built to handle interactions of varying strength, like ALS-WR. This again is in Mahout, and is what Myrrix is based on too.