Recommended Content Display Using Consumption Feature Labels

The patent segments users into different consumption types (e.g., intensive readers vs. extensive readers) and segments recommended content by consumption features. This segmentation allows the system to match content with appropriate user groups, maintaining recommendation accuracy while expanding the overall quantity of available content through targeted segmentation rather than homogeneous mixing.

The patent applies local quality by attaching consumption feature labels to specific pieces of recommended content based on the publisher's consumption characteristics. Different regions of the content ecosystem (different recommendation streams) have different quality characteristics matched to different user consumption types, ensuring that each user receives content with appropriate local quality properties.

The patent extracts essential consumption features from complex user behavior data and represents them as simplified labels attached to content. This extraction process takes out only the most relevant characteristics (e.g., consumption intensity, content type preference) from the full complexity of user behavior, maintaining discovery accuracy while reducing the complexity burden on the system.

The patent transforms complex user consumption behavior into discrete parameter labels that can be easily processed and matched. By changing the representation of consumption features from continuous complex data to discrete categorical parameters, the system achieves accurate content discovery with reduced computational complexity.