Viewport Tracking for Social Feed Relevance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional online social networking systems face challenges in accurately tracking user interactions with content items, leading to irrelevant content being displayed to users, which affects user engagement and the quality of the feed.
Innovation Solution
Viewport Tracking (VPT) is implemented on the client side to track what content items are actually rendered within the user's viewport, including the time of entry and duration, providing more accurate data on user interaction compared to server-side tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If server-side tracking is used to track content items, then implementation is simpler, but measurement precision of user interaction is insufficient
Solution Approach 1:
The patent introduces a client-side tracking mechanism as an intermediary between the server and user interaction. This intermediary (client-side code) captures actual viewport data and sends it to the server, resolving the contradiction by adding measurement precision without requiring complete system redesign. The intermediary handles the complexity locally while maintaining server-side simplicity.
Solution Approach 2:
The patent transitions from server-side tracking (one dimension) to client-side viewport tracking (another dimension). This dimensional shift enables precise measurement of actual user interaction by tracking what is visible in the user's viewport, thereby improving measurement precision while managing complexity through targeted implementation.
2Reliability
If conventional tracking methods are used, then system complexity is lower, but content relevance decreases
Solution Approach 1:
The patent implements feedback by tracking actual user interaction data (viewport information) and using this feedback to improve content ranking and relevance. The system continuously learns from user behavior patterns, adjusting content delivery to improve relevance. This feedback mechanism resolves the contradiction by making the system adaptive without requiring complete complexity overhaul.
Solution Approach 2:
The patent performs preliminary tracking of viewport data at the client side before content is fully processed server-side. This preliminary action captures essential interaction information early, allowing the system to pre-rank or pre-filter content for better relevance while maintaining manageable implementation complexity through incremental integration.
3Manufacturing precision
If more accurate viewport tracking is implemented, then content item ranking quality improves, but data processing requirements increase
Solution Approach 1:
The patent extracts only the essential viewport tracking data (entry time, duration, visibility status) from the full user interaction stream. By taking out only the critical precision metrics needed for ranking rather than all possible interaction data, the system achieves high ranking precision while minimizing data processing volume and storage requirements.
Solution Approach 2:
The patent applies partial tracking by focusing viewport monitoring on specific content items and interaction events rather than all possible user actions. This selective approach provides sufficient precision for content ranking without the excessive data processing burden of tracking every conceivable user interaction detail.
Data Source
AI summary
A system and method includes obtaining activity data of an online social networking system, the activity data including duration data indicative of a duration in which at least some of the content items has previously been displayed on other feeds of members of the online social networking system and interactions with the content items by the members. For each of the content items, a probability score of the member to interact with the content item is determined by applying a logistic regression model to the duration data and the interactions to determine, at least in part, a probability score of the member to interact with the content item and the feed is generated with at least some of the content items having an order based on the probability score for each of the at least some content items.