Cross-Domain Recommendation Model Training via Heterogeneous Semantic Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recommendation methods are affected by the scale of users, leading to insufficient correlation between recommended target content and preferred content, especially when the user data is sparse.
Innovation Solution
A cross-domain recommendation model training method that constructs a heterogeneous network with node bipartite graphs and label bipartite graphs, generating training samples based on source-domain and target-domain content nodes and their corresponding semantic labels, and training a cross-domain recommendation model using these samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If user nodes are introduced to the recommendation method, then the recommendation can be made according to user preferences, but the recommendation accuracy deteriorates when the user scale is small
Solution Approach 1:
The patent introduces content nodes as intermediaries between users and recommended content. Instead of directly modeling user preferences, the system uses content nodes with semantic labels to represent and transfer preference information across domains. This mediator approach allows the system to overcome user data sparsity by leveraging content-level semantic relationships.
Solution Approach 2:
The patent transitions from a traditional user-item recommendation space to a heterogeneous network space that includes user nodes, content nodes, and semantic label nodes. This dimensional expansion creates additional paths for preference propagation, particularly through the content nodes and their semantic labels, enabling accurate recommendations even with limited user data.
2Productivity
If traditional recommendation methods are used with limited user data, then the system can operate with small user scale, but the correlation between recommended content and user preference becomes insufficient
Solution Approach 1:
The patent pre-constructs a heterogeneous network that includes content nodes and semantic label nodes before the recommendation process. This preliminary action creates a rich structural framework with pre-established semantic relationships, allowing the system to immediately leverage content-level information without requiring extensive user interaction data.
Solution Approach 2:
The content nodes in the heterogeneous network serve multiple functions: they represent actual content, encode semantic information through labels, and act as aggregation points for user preferences. This multi-functionality allows the system to extract meaningful signals from limited user data by processing them through the versatile content node structure.
Data Source
AI summary
A cross-domain recommendation model training method includes: constructing a heterogeneous network, the heterogeneous network including a node bipartite graph between sample source-domain content nodes and sample target-domain content nodes, a first label bipartite graph between the sample source-domain content nodes and sample source-domain semantic labels, and a second label bipartite graph between the sample target-domain content node and sample target-domain semantic labels; generating a training sample based on a sample source-domain content node and a sample target-domain content node between which a connecting edge exists in the node bipartite graph, a sample source-domain semantic label corresponding to the sample source-domain content node in the first label bipartite graph, and a sample target-domain semantic label corresponding to the sample target-domain content node in the second label bipartite graph; and training a cross-domain recommendation model based on the training sample.