The document describes combining content-based and collaborative filtering methods for information filtering. It discusses:
1) Content-based filtering recommends documents based on keywords in user profiles matching document content, while collaborative filtering leverages ratings from similar users.
2) The combined method computes estimates using content-based filtering, then finds similar users to make recommendations using both ratings and estimates, weighted by similarity.
3) Experiments on movie rating datasets evaluate the methods based on coverage, accuracy, and other metrics, finding the combined method performs best.