Collaborative-Filtered Content Recommendations With Justification in Real-Time

A system comprising:

• an affinity module implemented at least partially in hardware of at least one computing device to determine user affinity scores for a plurality of attributes of digital content;
• a collaborative filtering module implemented at least partially in the hardware of the at least one computing device to filter digital content items of a catalog for recommendation by:
• identifying a first set of the digital content items that are associated with a first attribute of the plurality of attributes, the first attribute used for the identifying being based on the user affinity scores;
• ranking the first set of digital content items based on the user affinity scores and associations of the first set of digital content items with the plurality of attributes;
• and
• selecting a subset of digital content items from the first set of digital content items based on the ranking to form at least one recommendation for the subset of digital content items.

In a digital medium environment to provide digital content recommendations to client device users, a method implemented by a computing device, the method comprising:

• generating, by the computing device, a user-item matrix having cells that indicate strength-of-interaction measures of a client device user with a catalog of digital content items;
• generating, by the computing device, an item-attribute matrix having cells that indicate associations of digital content items of the catalog with a plurality of attributes of digital content;
• computing, by the computing device, user affinity scores based on the user-item matrix and the item-attribute matrix;
• identifying, by the computing device and from the catalog, a first set of digital content items associated with a first attribute of the plurality of attributes, the first attribute used for the identifying being based on the user affinity scores;
• determining, by the computing device, a latent factor vector for the client device user based on the user affinity scores;
• extracting, by the computing device, latent factor vectors for the first set of digital content items from the user-item matrix using matrix factorization;
• computing, by the computing device, dot products of the latent factor vector of the client device user and the latent factor vectors of the first set of digital content items;
• selecting, by the computing device, a subset of digital content items from the first set of digital content items based on the dot products;
• and
• providing, by the computing device and to the client device, at least one recommendation for the subset of digital content items.