Cold-Start Recommendation Graphs Combining Semantic and Co-View Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current content recommendation systems rely on user interactions to identify items, limiting recommendations to a small percentage of items with high interaction rates, failing to surface items with low or no prior user interaction.
Innovation Solution
A graph-based cold-start model combines a semantic similarity component and a viewed-also-viewed component to select interface items, generating a sparse distance matrix and a co-viewed weighted matrix, which are combined into an adjacency matrix to identify candidate items, ranked, and included in network interfaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current recommendation systems rely on user interactions to identify recommended items, then the system can identify items with high interaction rates, but the system limits recommendations to a small percentage of items and fails to surface items with low or no prior user interaction
Solution Approach 1:
The system pre-computes graph embeddings and adjacency matrices for all items in the catalog before recommendations are needed. This preliminary action creates a ready-to-use graph structure that enables fast retrieval and ranking of both popular and obscure items, eliminating the need to wait for user interactions to build recommendation data
Solution Approach 2:
The patent introduces a graph-based embedding model as an intermediary between user interactions and item recommendations. This model creates vector representations of items and computes similarity metrics that serve as a mediator to surface cold-start items alongside popular items, bridging the gap between interaction-based and content-based recommendation approaches
2Reliability
If the system surfaces only items with high interaction rates, then the recommendation accuracy is high, but the diversity of recommendations is limited
Solution Approach 1:
The system applies different weighting strategies to different items based on their characteristics. Popular items are ranked by interaction strength, while cold-start items are boosted through graph embedding similarity scores. This local quality adjustment allows the system to maintain high accuracy for known items while diversifying recommendations by promoting underutilized items that are semantically similar to user preferences
Solution Approach 2:
The patent combines multiple recommendation signals into a composite ranking system. The final recommendation score integrates both interaction-based metrics (for reliability) and graph embedding-based similarity metrics (for diversity). This composite approach merges the strengths of collaborative filtering and content-based recommendation to achieve both accuracy and diversity
Data Source
AI summary
Systems and methods of generating interfaces including recommended items selected by a graph-based cold-start (GCS) model are disclosed. A request for an interface is received and a set of interface items is generated for inclusion in the interface. The set of interface items is selected, at least in part, by a GCS model including a semantic similarity component and a viewed-also-viewed component. The set of interface items is generated based on a combination of an output of the semantic similarity component and an output of the viewed-also-viewed component. The interface including the set of interface items is generated and transmitted to a system that generated the request for the interface.