Search Ranking with Multi-Objective Searchee Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search engine ranking systems prioritize searcher-centric approaches, leading to sub-optimal results for searchee users, as they do not incorporate information about downstream interactions of searchee users, resulting in biased rankings that may not favor content creators with smaller audiences.
Innovation Solution
A searchee-searcher optimization system that utilizes multi-task learning and multi-objective optimization to generate outcome predictions for both searcher and searchee objectives, combining these predictions into a ranking score to optimize search result rankings for multiple conflicting objectives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If searcher-centric ranking approaches are used, then searcher user experience is improved, but searchee user engagement deteriorates
Solution Approach 1:
The patent segments the ranking objective into two distinct components: searcher-centric objectives (query satisfaction, relevance) and searchee-centric objectives (content creator engagement, visibility). By separating these objectives and optimizing for both simultaneously, the system resolves the contradiction between improving searcher experience while maintaining searchee engagement.
Solution Approach 2:
The patent introduces a new dimension to the ranking problem by incorporating searchee user characteristics and downstream interaction potential as additional ranking signals. This multi-dimensional approach allows the system to balance searcher satisfaction with searchee engagement, moving beyond the traditional single-dimension searcher-centric model.
2Productivity
If traditional ranking models are used, then computational efficiency is maintained, but ranking fairness deteriorates
Solution Approach 1:
The patent introduces an intermediary component that bridges traditional ranking models and fairness considerations. This intermediary layer processes additional signals about searchee user characteristics and downstream interaction potential, integrating fairness constraints into the existing ranking pipeline without completely replacing the efficient traditional models.
Solution Approach 2:
The patent modifies ranking parameters by incorporating new features related to searchee user characteristics, audience size, and downstream interaction potential. These parameter changes enable the system to achieve better fairness while maintaining computational efficiency through optimized feature processing and model architecture.
3Ease of operation
If searcher-only objectives are optimized, then search query satisfaction is improved, but content creator visibility deteriorates
Solution Approach 1:
The patent merges searcher-centric objectives with searchee-centric objectives into a unified ranking framework. By combining these previously separate optimization goals, the system simultaneously improves query satisfaction and content creator visibility, preventing the loss of information about creator engagement in the ranking process.
Solution Approach 2:
The patent creates a multi-functional ranking system that serves multiple purposes: satisfying searcher queries, promoting content creator visibility, and encouraging downstream user interactions. This universal approach allows the ranking model to fulfill multiple functions simultaneously, rather than prioritizing only query satisfaction.
Data Source
AI summary
Embodiments include technologies to apply at least one machine learning model to features of a search query, features of a searcher user, features of a searchee content item, and features of a searchee user, produce a first outcome prediction that represents a probability of a first objective relating to engagement of the searcher user with a content item in an online system and a second outcome prediction that represents a probability of a second objective relating to engagement of the searchee user with the online system responsive to the engagement of the searcher user with the content item, apply a multi-objective optimization solver to the first objective, the second objective and an outcome prediction that is a combination of the first outcome prediction and the second outcome prediction, and generate a serving function for a search engine based on the first objective, the second objective, and the outcome prediction.