Two-Stage Job Ranking Using Global and User-Specific Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Scalable response prediction in large data sets is hindered by high network communication costs and latency due to the need for processing millions of jobs with separate sets of coefficients in conventional recommendation systems.
Innovation Solution
A system utilizing a combination of global and user-specific statistical models, with a two-stage ranking process that applies a global model to generate an initial ranking and then uses user-specific and job-specific models to refine recommendations, while reducing network communication costs by grouping feature vectors and using an inverted index for efficient data retrieval.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If separate sets of coefficients are used for each user in conventional recommendation systems, then personalized recommendation accuracy is improved, but network communication costs and processing latency increase significantly
Solution Approach 1:
The patent segments the coefficient storage and processing into user-specific portions and global portions. Each user has their own coefficient set for personalized recommendations, while shared global coefficients are maintained separately. This segmentation allows the system to retrieve only necessary coefficient portions during inference, reducing network communication overhead and processing latency while maintaining personalized recommendation accuracy.
2Measurement precision
If millions of jobs are processed with user-specific models, then recommendation personalization is improved, but system scalability deteriorates
Solution Approach 1:
The patent merges user-specific coefficient processing with global coefficient processing into a unified model structure. By combining these previously separate processing paths, the system achieves better resource utilization and faster processing. The merged approach allows parallel computation of user-specific and global contributions, improving system scalability while maintaining personalized recommendation quality.
3Ease of operation
If conventional software tools are used to handle large data sets, then ease of operation is maintained, but the ability to process big data deteriorates
Solution Approach 1:
The patent introduces an intermediary layer (the specialized recommendation system with optimized coefficient storage and retrieval mechanisms) that sits between conventional software tools and the big data. This intermediary handles the complexity of processing millions of user profiles and job listings, translating large-scale data operations into efficient computations. Users can operate the system through familiar interfaces while the intermediary manages the underlying big data processing challenges.
Data Source
AI summary
The disclosed embodiments provide a system for processing data. During operation, the system obtains a global version of a statistical model and a user-specific version of the statistical model for a user. Next, the system applies the global version to member features of the user and job features of a set of jobs to generate a first ranking of the jobs for the user. The system then applies the user-specific version to the member features and the job features for a highest-ranked subset of jobs in the first ranking to generate a second ranking of the jobs for the user. Finally, the system outputs at least a portion of the second ranking as a set of job recommendations.