Followfeed Impression Discounting for Fresher Content Ranking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing online connection network systems perform impression discounting at a final stage, leading to inefficient use of bandwidth and degrading user experience due to the presence of older content in user feeds, despite increasing uniqueness of results.
Innovation Solution
Implementing impression discounting at the candidate generation layer and applying a global freshness factor to decay updates based on age buckets, ensuring fresher content is prioritized while maintaining user engagement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If impression discounting is performed at the final stage, then the system can ensure uniqueness of results by filtering previously seen content, but it wastes valuable time and bandwidth retrieving content that will not be of interest to the user
Solution Approach 1:
The patent applies impression discounting at the candidate generation layer rather than at the final stage. This preliminary action filters out previously seen content before the ranking and retrieval processes, preventing waste of time and bandwidth on content that would be discarded anyway. The system maintains uniqueness by establishing the discounting mechanism early in the content delivery pipeline.
2Device complexity
If traditional filtering is applied only at the final stage, then implementation is simpler, but older content appears in user feeds degrading user experience
Solution Approach 1:
The patent segments the content filtering process into multiple stages: candidate generation layer (first pass) and final ranking stage (second pass). The candidate generation layer applies initial impression discounting to eliminate old content early, while the second pass performs final filtering. This segmentation improves user experience by ensuring fresher content reaches the feed while distributing the computational complexity across different layers.
Solution Approach 2:
By applying impression discounting and freshness factors at the candidate generation layer before final ranking, the system preliminarily filters out older content. This preliminary action ensures that only potentially relevant, fresher content proceeds to the final ranking stage, significantly improving user experience without overwhelming complexity at any single stage.
3Adaptability or versatility
If the system retrieves more unique content, then it increases the variety of feed content, but it includes older content that reduces freshness perception
Solution Approach 1:
The patent applies different quality criteria to different stages of content processing. At the candidate generation layer, the system prioritizes freshness by applying age-based discounting and freshness factors. At the final ranking stage, it focuses on relevance and engagement. This local quality approach ensures that freshness is maintained in the candidate pool while variety is optimized in the final feed composition.
Solution Approach 2:
The system dynamically adjusts the impression discounting parameter based on content age. Fresher content receives lower discounting (higher priority), while older content receives higher discounting (lower priority). This parameter change mechanism allows the system to balance between content variety and freshness, ensuring that the feed adapts to user preferences for current, relevant content.
Data Source
AI summary
Described herein is a technique for surfacing content for users of a connection network application. The technique involves performing a first query to fetch recently-impressed items viewed by a user of the connection network application and, concurrently with the first query, performing a second query of search nodes to generate a set of search node results. The technique allows for filtering the recently-impressed items from the set of search node results to generate a candidate set and applying a freshness factor to the candidate set. Updates are provided to a user display, based on results of applying the freshness factor to the candidate set.