Ephemeral Content Ranking via Selection and Time Probability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional social networking systems fail to effectively rank ephemeral content item collections in a way that maximizes user engagement, as chronological or reverse chronological ordering does not guarantee that users will find the most interesting content.
Innovation Solution
The system ranks ephemeral content item collections based on a probability of user selection and time spent on each collection, using machine learning models trained on attributes such as user behavior, content characteristics, and collection features, to prioritize content that users are likely to engage with.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If chronological or reverse chronological ordering is used to rank ephemeral content item collections, then the system maintains simplicity in ranking logic, but user engagement is not maximized as users may not find the most interesting content
Solution Approach 1:
The patent changes the ranking parameters from simple temporal metrics (chronological order) to a composite scoring system that incorporates multiple factors including probability of user selection, probability of time spent, and recency. This transforms the ranking mechanism from a single-dimensional temporal sort to a multi-dimensional evaluation, thereby improving user engagement while maintaining computational feasibility through structured parameter integration.
2Measurement precision
If machine learning models are trained on multiple features including user behavior and content characteristics, then the accuracy of content ranking is improved, but the system complexity increases
Solution Approach 1:
The patent segments the complex ranking problem into distinct components by training separate machine learning models for different aspects: one model predicts probability of user selection, another predicts probability of time spent, and a third handles recency. Each model focuses on specific features relevant to its prediction task, making the overall system more manageable and interpretable while achieving high ranking accuracy through coordinated use of these specialized models.
3Productivity
If the system prioritizes content based on probability of user selection and time spent, then user engagement is enhanced, but the computational resources required for real-time scoring increase
Solution Approach 1:
The patent performs preliminary actions by pre-training machine learning models offline using historical data, and by pre-computing feature representations of content items and user profiles. During real-time operation, the system only needs to apply these pre-trained models to current inputs and combine pre-computed features with live user interactions, significantly reducing the computational burden during peak usage periods while maintaining accurate, engagement-optimized ranking.
Data Source
AI summary
Systems, methods, and non-transitory computer readable media can perform a first ranking to rank each ephemeral content item collection of a plurality of ephemeral content item collections based on a probability of a user selecting the ephemeral content item collection, each ephemeral content item collection of the plurality of ephemeral content item collections including one or more ephemeral content items. One or more ephemeral content item collections from the first ranking can be provided in an ephemeral content feed of the user. A selection by the user of an ephemeral content item collection provided in the ephemeral content feed can be received. A second ranking to rank each ephemeral content item collection of the ephemeral content item collections provided in the ephemeral content feed other than the selected ephemeral content item collection based on a probability of the user spending time on the ephemeral content item collection can be performed.