Attribute-Based Session Ranking With Real-Time Affinity Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ranking systems on exchange platforms are not responsive enough to a user's changing preferences within a web session, as they are typically trained on vast historical data spanning months or years, failing to adapt quickly to real-time user interactions.
Innovation Solution
An attribute-based ranking system that determines user affinity scores for item attributes using a multi-armed bandit model, updating an attribute repository in real-time based on user interactions during a session, and re-ranking items accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing ranking systems are trained on vast historical data spanning months or years, then ranking accuracy is improved, but responsiveness to user's changing preferences within a web session deteriorates
Solution Approach 1:
The patent segments the ranking system into two distinct components: a pre-trained ranking model that handles general ranking accuracy using historical data, and a session-aware bandit model that handles real-time adaptation within web sessions. This segmentation allows each component to specialize in one aspect, resolving the contradiction between historical accuracy and real-time responsiveness.
Solution Approach 2:
The patent introduces a dynamic session-aware bandit model that continuously updates user preference estimates during the web session based on observed interactions. This dynamic component adapts to changing user preferences in real-time, complementing the static pre-trained model and enabling responsiveness without sacrificing overall ranking accuracy.
2Adaptability or versatility
If complex ranking models are used to capture user preferences, then personalization quality is improved, but computational complexity and resource usage increase
Solution Approach 1:
The patent employs a lightweight bandit model that requires minimal computational resources compared to complex deep learning models. This simpler model can be rapidly updated during each web session without significant computational overhead, enabling personalization while keeping device complexity manageable.
Solution Approach 2:
The patent changes the approach from learning complex user profiles to estimating simple preference parameters using bandit algorithms. By focusing on parameter estimation rather than full profile learning, the system achieves personalization with reduced computational complexity and faster update times during web sessions.
Data Source
AI summary
Methods, systems, and apparatuses, including computer programs encoded on a computer storage medium, that rank items based on user interactions within the same web session. The method includes: providing a first content page for display on a client device; receiving a set of user interactions with one or more items of the first plurality of items; determining an affinity score representing a user interest in the attribute based on the set of user interactions; updating an attribute repository storing a second plurality of attributes and corresponding current affinity scores; identifying a second plurality of items and a corresponding third plurality of attributes; ranking the second plurality of items; and providing a second content page for display on the client device.