Affinity Scoring for Heterogeneous Content Feeds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Social networking services face challenges in providing personalized and relevant content to users, as existing algorithms struggle to accurately identify user affinity for articles and topics, leading to a heterogeneous feed that may not effectively capture user interests.
Innovation Solution
The method involves retrieving user and article features, determining a score based on these features using vector calculations, and ranking articles for presentation in a user's feed, while also identifying topics of interest and providing recommendations based on the affinity score, with offline and online scoring components optimized for computational efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing algorithms are used to identify user affinity for articles, then the system can provide content recommendations, but the accuracy of identifying user affinity is insufficient
Solution Approach 1:
The patent segments the affinity scoring process into two distinct components: offline scoring (pre-computing user features, article features, and base affinity scores) and online scoring (computing real-time affinity adjustments based on user interactions). This segmentation allows each component to be optimized independently, improving overall measurement precision without compromising reliability
Solution Approach 2:
The system performs preliminary actions by pre-computing user features, article features, and base affinity scores during offline processing. This preliminary action reduces the computational burden during online scoring and improves the accuracy of real-time recommendations by having pre-processed data ready for quick retrieval and combination
2Adaptability or versatility
If the system provides a heterogeneous feed of content, then users receive diverse content types, but the feed may not effectively capture user interests
Solution Approach 1:
The patent applies local quality by computing affinity scores specifically for articles within the heterogeneous feed, rather than treating all content types uniformly. The system extracts article-specific features and computes affinity scores tailored to each article, ensuring that the diversity of content types is effectively captured and personalized for each user
Solution Approach 2:
The system changes parameters by computing affinity scores based on multiple features including user features, article features, and interaction features. The affinity score is dynamically adjusted based on these parameter changes, allowing the system to effectively capture user interests across diverse content types in the heterogeneous feed
3Reliability
If the system computes affinity scores for all articles, then comprehensive recommendations are provided, but computational resources are consumed
Solution Approach 1:
The patent segments the scoring process into offline and online components to improve computational efficiency. Offline scoring pre-computes user features, article features, and base affinity scores, while online scoring only computes real-time affinity adjustments. This segmentation maintains comprehensive recommendation quality while significantly reducing real-time computational resource consumption
Solution Approach 2:
The system performs preliminary affinity scoring computations during offline processing, storing pre-computed scores and features for quick retrieval during online operations. This preliminary action ensures comprehensive recommendations are available while minimizing real-time computational overhead, thereby improving productivity without sacrificing reliability
Data Source
AI summary
The present disclosure describes various embodiments of methods, systems, and machine-readable mediums which help determine a user's likely affinity for an article presented (or to be presented) in a heterogeneous feed of a social network.