Spaced Recommendations via Repulsion Force Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recommendation algorithms in online environments often fail to provide diverse and interesting recommendations, leading to user dissatisfaction and reduced sales revenue, as they typically focus on obvious or homogeneous suggestions that do not impart new information to users.
Innovation Solution
The implementation of 'spaced recommendations' generated through 'repulsion force' analysis, which separates recommendations from each other and from known user interests, ensuring a diverse and novel set of suggestions by applying algorithms that either mutually space recommendations or exclude obvious items, thereby reducing clustering around a single theme.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the recommendation engine generates recommendations based on user purchase history and interest assessment, then the recommendations match user interests and increase relevance, but the recommendations become homogeneous and fail to provide novel information
Solution Approach 1:
Instead of only selecting recommendations closest to user interests, the patent inverts the approach by deliberately selecting recommendations at varying distances from the core interest cluster. The system ranks recommendations by distance metric and selects a diverse subset that includes both highly relevant and somewhat distant items, thereby providing novel information while maintaining relevance.
Solution Approach 2:
The patent changes the selection parameter from purely similarity-based to distance-based ranking. By computing distance metrics between recommendation items and user interest profiles, and selecting items based on their distance rankings rather than similarity scores, the system transforms the recommendation selection process to achieve both relevance and novelty.
2Reliability
If the recommendation engine selects only the most relevant recommendations, then the recommendations closely match user interests, but the recommendations cluster around a single theme and reduce user engagement
Solution Approach 1:
The patent segments the recommendation selection process into multiple stages: generating candidate recommendations, ranking them by distance metric, and selecting a diverse subset. This segmentation allows the system to maintain accuracy by considering relevant items while achieving diversity through the distance-based selection criterion that prevents over-concentration on a single theme.
Solution Approach 2:
The patent adds a distance dimension to the recommendation selection process. By computing and utilizing distance metrics between recommendations and user interests as a selection criterion, the system moves beyond traditional similarity-based selection to a multi-dimensional approach that simultaneously achieves accuracy and diversity.
3Measurement precision
If the recommendation engine provides highly specific recommendations based on user behavior patterns, then the recommendations are precisely targeted, but the recommendations become obvious and fail to interest users
Solution Approach 1:
The patent inverts the traditional approach by selecting recommendations based on their distance from user interest profiles rather than their similarity. This inversion ensures that selected recommendations include items that are precisely targeted (based on distance calculation) but also interesting (because they are not obvious repetitions of user's known interests).
Data Source
AI summary
Strategies are described for generating recommendations. The strategies generate a set of original recommendations based on a source of information. The strategies then transform the set of original recommendations into a set of similarity-spaced recommendations based on “repulsion force” analysis applied to the set of original recommendations. In a first implementation, the set of spaced recommendations represent a diverse sampling of items in the set of original recommendations. In a second implementation, the set of spaced recommendations represent a sampling of items in the set of original recommendations which omits or excludes recommendations assessed as obvious. A third implementation can combine the first and second implementations.