?

Log in

fictionmenu in fiction_menu

Recommender System

 


Now I'm thinking about the real relevance of fictionmenu recommender system. Today's most common recommender algorithm is to select users who liked the same items as you did, and recommend the items they liked. This is how fictionmenu system works today. The idea is that users with similar sets of favorite books/movies have got similar personal characteristics that predictively determine their further choices. It seems to be expedient but I think there's a problem. Any great book or movie comprises a semantic universe where every intelligent person can find very individual things. And if we scan the brains of two different guys who have read the same book, we'll possibly find out that they've read different texts. Even if they both liked the book.

There's another aspect of the problem of the recommender system. Do people follow these recommendations? What if their friends off-line are the only trusted source for book/movie recommendations? Maybe it's more important to maintain search mechanism and browsing interface?


What do you think about it?

 

Comments

logo

September 2008

S M T W T F S
 123456
78910111213
14151617181920
21222324252627
282930    
Powered by LiveJournal.com