1 Assignment Instructions

Now that we have covered basic techniques for recommender systems, choose one commercial recommender and describe how you think it works (content-based, collaborative filtering, etc). Does the technique deliver a good experience or are the recommendations off-target?

You may also choose one of the three non-personalized recommenders (below) we went over in class and describe the technique and which of the three you prefer to use.

  1. Metacritic: How We Create the Metascore Magic
  2. Rotten Tomatoes: About Rotten Tomatoes
  3. IMDB: FAQ for IMDb Ratings

2 IMDB Recommender System

IMDB uses following hybrid recommendation systems by which system tries to find the links between contents through observing users’ patterns of watching films and reviews : - content-based recommendation systems which suggests movies as per “People who liked this also liked”. - tag-based systems where IMDb can provide certain tags (genres of movies), and users can tag content themselves. (A Netflix Prize winning model called latent factor model is tag-based)

IMDB primarily uses Graph Database to recommend movies and maintain feedback

Accordingly, for a given user, the recommendation is performed in three steps:

  • Identification of movies rated above a predefined appreciated threshold,
  • Identification of unwatched movies similar to the appreciated ones with similarity score above a predefined similarity threshold,
  • For movies similar to several appreciated ones, computation of the sum of the similarity scores.

3 Challenges

  • To prevent systems from being too convergent and recommend similar content all the time, and eventually bore the users
  • Fix the cold start problem - there needs to be enough other users already in the system to find a match.
  • Sparsity : If there are many items to be recommended, even if there are many users, the user/ratings matrix is sparse, and it is hard to find users that have rated the same items
  • First Rater : Cannot recommend an item that has not been previously rated (New items and Esoteric items)
  • Popularity Bias : Cannot recommend items to someone with unique tastes - Tends to recommend popular items.

4 Possible Approach

One of the key metrics used in outcome assessment is CTR (click through rate). For example, when we recommend a movie to users, 10,000 users may/may-not see the movie feed, but it is the percentage of users who click the link and actually look at the content that determines how effective the recommendation is. Theoretically, CTR is higher when the quality of recommendation is better.

To prevent over-sensationalizing headlines affecting the result, we could use other metrics based on the interaction between users and content, such as the number of “like”, number of comments and time spent reading (watching, listening to) it. We could label each action with a number, for example, click through=1, like=2, comment=3 and share=4, this allows us to calculate the similarity of two things using cosine similarity, and this is also useful when trying to determine how strong a tie is between users