Interactive Recommendation System with Segmented User History
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recommendation systems are not well-suited for tracking changes in user interest in real-time, leading to delayed updates and inefficient use of network and computation resources, resulting in poor user experience.
Innovation Solution
An interactive recommendation system that includes an information generation module to provide and update content based on user selections, using a machine learning algorithm to predict user interest and rank information items, allowing for timely alignment with user preferences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the recommendation system uses accumulative user behavior history to predict user interest, then the recommendation accuracy over long period is improved, but the response time to track user interest changes deteriorates
Solution Approach 1:
The patent segments user behavior history into two distinct components: recent behavior history (short-term) and long-term behavior history. This segmentation allows the system to process and respond to recent changes quickly while maintaining the benefit of long-term patterns for accurate recommendations, thereby resolving the contradiction between response time and recommendation accuracy.
Solution Approach 2:
The patent introduces a dynamic adjustment mechanism where the weight assigned to recent behavior history versus long-term behavior history can be adjusted based on user characteristics and context. This dynamic weighting allows the system to prioritize recent changes when needed for timely responses, while maintaining long-term accuracy when appropriate, thus resolving the time-response contradiction.
2Measurement precision
If the recommendation system continuously updates recommendations based on user behavior, then the relevance of recommendations is improved, but the network and computation resource consumption increases
Solution Approach 1:
Instead of continuous updates, the patent implements periodic or event-triggered updates based on user behavior. The system updates recommendations when significant user interactions occur (such as viewing, clicking, or completing actions on recommended items) rather than continuously monitoring and updating, thereby reducing network and computation resource consumption while maintaining high recommendation relevance.
Solution Approach 2:
The system uses user interactions themselves as the trigger for updates, where user behavior automatically initiates the recommendation update process without requiring continuous system-initiated queries or checks. This self-service approach reduces unnecessary network requests and computation resources while ensuring recommendations remain relevant to current user interests.
3Measurement precision
If the recommendation system waits for sufficient accumulative data to manifest interest changes, then the accuracy of interest detection is improved, but the time delay before updating recommendations deteriorates
Solution Approach 1:
The patent segments user behavior data into recent and long-term components, allowing the system to detect interest changes in the recent segment quickly without waiting for sufficient accumulative data from the long-term segment. This enables timely detection of emerging interests while maintaining the contextual accuracy provided by long-term behavior patterns.
Solution Approach 2:
The system performs preliminary analysis of recent user behavior patterns to detect potential interest changes before they fully manifest in accumulative data. By analyzing recent behaviors (such as sequential views, clicks, or interactions) as they occur, the system can proactively update recommendations in advance, reducing time delay while maintaining detection accuracy.
Data Source
AI summary
An interactive recommendation system is provided. The system comprises an information generation module configured to: provide a set of information contents for outputting in an information interface, wherein each of the set of information contents is associated with an information item; receive a selection of one of the set of information contents that signals the user's interest or disinterest in an information item; determine, based on a model and history data related to prior activities of the user including a plurality of selections from the set of information contents, one or more information items to be included in recommendation information for the user; provide the recommendation information for outputting in a recommendation information interface; and update, based on the history data, the set of information contents for outputting in the information interface.