Candidate Generation

Generate movie candidates using shared viewing histories, content attributes, and learned embeddings. Compare what evidence each method needs and how that affects new users and newly added titles.

In this lesson, we will be looking at a few techniques to generate media candidates that will match user interests based on the user’s historical interaction with the system.

Candidate generation techniques

The candidate generation techniques are as follows:

1. Collaborative filtering

2. Content-based filtering

3. Embedding-based similarity

Combine candidates from these sources before passing them to the ranker. A title can be retrieved because similar users watched it, because its metadata matches prior watches, or because it scores highly in the learned embedding space.

Collaborative filtering

In this technique, you find users similar to the active user based on the intersection of their historical watches. You, then, collaborate with similar users to generate candidate media for the active user, as shown below.