I have implemented such an algorithm in the past. The problem is that it actually makes things even more redundant since it gives you stuff you have already read and liked. The results returned are usually articles on the exact same topic from a different source.
