US Patent Application 18353789. MULTISTAGE FEED RANKING SYSTEM WITH METHODOLOGY PROVIDING SCALABLE MULTI-OBJECTIVE MODEL APPROXIMATION simplified abstract
MULTISTAGE FEED RANKING SYSTEM WITH METHODOLOGY PROVIDING SCALABLE MULTI-OBJECTIVE MODEL APPROXIMATION
Organization Name
Microsoft Technology Licensing, LLC
Inventor(s)
Madhulekha Arunmozhi of Sunnyvale CA (US)
Ian Ackerman of Mountain View CA (US)
Manas Somaiya of Sunnyvale CA (US)
MULTISTAGE FEED RANKING SYSTEM WITH METHODOLOGY PROVIDING SCALABLE MULTI-OBJECTIVE MODEL APPROXIMATION - A simplified explanation of the abstract
This abstract first appeared for US patent application 18353789 titled 'MULTISTAGE FEED RANKING SYSTEM WITH METHODOLOGY PROVIDING SCALABLE MULTI-OBJECTIVE MODEL APPROXIMATION
Simplified Explanation
- This patent application describes techniques for approximating a complex multi-objective feed item scoring model with a simpler single objective model in a multistage feed ranking system. - The goal is to optimize the personalization and ranking of feeds by considering factors such as viewer experience, professional or social network effects, and content creator impact. - The techniques involve using a lightweight single objective model at the first pass ranker stage, which efficiently scores a large set of feed items. - Despite being less complex, the single objective model still maintains much of the richness and complexity of the multi-objective model. - The second pass ranker stage uses the more complex multi-objective model, but with the benefit of high recall due to the efficient scoring done by the first pass ranker. - Overall, these techniques enable effective multi-objective optimization in the ranking system while improving efficiency and maintaining a high level of performance.
Original Abstract Submitted
Approximating a more complex multi-objective feed item scoring model using a less complex single objective feed item scoring model in a multistage feed ranking system of an online service. The disclosed techniques can facilitate multi-objective optimization for personalizing and ranking feeds including balancing personalizing a feed for viewer experience, downstream professional or social network effects, and upstream effects on content creators. The techniques can approximate the multi-objective model—that uses a rich set of machine learning features for scoring feed items at a second pass ranker in the ranking system—with the more lightweight, single objective model—that uses fewer machine learning features at a first pass ranker in the ranking system. The single objective model can more efficiently score a large set of feed items while maintaining much of the multi-objective model's richness and complexity and with high recall at the second pass ranking stage.