How a Recommendation System Uses Dislikes to Suggest New Content

Comcast's patent describes a content recommendation system that finds users with similar tastes by focusing on what content they both dislike, especially popular disliked items, to suggest new shows or movies.

What it covers

This patent describes a system that recommends content by finding users who share similar dislikes. First, it identifies a "target user" and a group of "candidate users." For each candidate user, the system determines how many content items both the candidate and the target user have disliked (Claim 1). The system then ranks these candidate users, giving more importance to commonly-disliked items that are popular among other users (Claim 1, 3, 5, 6). For instance, if you and another user both disliked a widely popular movie, that shared dislike would count more towards finding you a good recommendation than if you both disliked a very obscure film. Finally, the system suggests new content to the target user based on what the highest-ranked, similarly-disliking candidate users have consumed, ensuring the target user has not seen it before (Claim 4).

What it doesn't cover

• —Recommendation systems that only consider content a user has liked or consumed, without factoring in dislikes.
• —Systems that do not give preference to commonly-disliked content items based on their overall popularity.
• —Recommendation methods that rely solely on explicit user ratings without comparing shared disliked items.
• —Systems that use a simple count of shared disliked items without weighting them by how popular those items are.
• —Recommendations based purely on content metadata or genre similarity without user preference analysis.

The clever bit

The clever part is identifying user similarity not just by what they like or consume, but by what they *dislike*, and specifically, giving more weight to shared dislikes of *popular* content. This helps the system find more meaningful connections between users by focusing on strong, shared negative opinions about mainstream content.

Why it matters

Content recommendation systems are crucial for modern media platforms, helping users discover new content and keeping them engaged. This patent's approach to using shared dislikes, particularly popular ones, offers a refined way to identify user preferences. It aims to improve the accuracy of recommendations, which can increase user satisfaction and content consumption on platforms like those offered by Comcast.

Real-world examples

• 1.Netflix recommendation engine
• 2.YouTube video suggestions
• 3.Amazon Prime Video recommendations
• 4.Hulu personalized content feeds
• 5.Spotify music discovery algorithms
• 6.Comcast Xfinity content recommendations