Explainable Embedding Recommendations Using Interpretation Hierarchies
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recommendation systems are considered black boxes, making it difficult to understand why certain recommendation results are generated, which is crucial for stakeholders such as users, engineers, and other business entities.
Innovation Solution
An explainable embedding-based recommendation system is developed using graph convolution operations to generate embeddings by disentangling information from different layers, allowing for the interpretation of user and item interactions through an interpretation hierarchy and knowledge graph entities, providing transparent recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If embedding techniques are used to improve recommendation performance, then recommendation accuracy is improved, but interpretability deteriorates (black box problem)
Solution Approach 1:
The patent introduces an interpretation hierarchy as an intermediary layer between the embedding-based recommendation model and the final output. This hierarchy includes multiple levels (entity level, category level, attribute level) that progressively interpret the embeddings, making the black-box recommendations transparent and explainable while preserving the accuracy benefits of embedding techniques
Solution Approach 2:
The patent segments the recommendation explanation into multiple hierarchical levels: entity-level interpretations (specific items), category-level interpretations (groups of items), and attribute-level interpretations (features and properties). This segmentation allows the system to provide comprehensive explanations at different granularities, resolving the interpretability issue while maintaining recommendation accuracy
2Reliability
If complex embedding models are used to achieve better performance, then recommendation quality is improved, but model complexity increases
Solution Approach 1:
The patent segments the complex embedding model into manageable hierarchical components (entity level, category level, attribute level), each handling specific aspects of the recommendation task. This segmentation reduces the apparent complexity by organizing the model structure systematically while maintaining overall recommendation quality
Solution Approach 2:
The patent adds a hierarchical interpretation dimension to the traditional flat embedding model. By introducing multiple levels of interpretation (from specific entities to general categories to abstract attributes), the system manages model complexity through dimensional organization without sacrificing recommendation quality
Data Source
AI summary
A method includes obtaining, by an electronic device, an interpretation hierarchy generated based on a knowledge graph and behavioral data. The method also includes performing, by the electronic device, graph convolution operations on the interpretation hierarchy to generate one or more embeddings. The method further includes generating, by the electronic device, a recommendation based at least in part on associations between the one or more embeddings. In addition, the method includes providing, by the electronic device, an explanation corresponding to the recommendation.