Explainable Knowledge Graph Embedding via Feature Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning methods lack the ability to generate interpretable explanations for knowledge graph embeddings, making it difficult to understand the reasoning behind predictions and decisions in technical systems.
Innovation Solution
A computer-implemented method and device that determine a set of features for entities in a knowledge graph, assigning scores to these features based on their relevance to the embedding, and selecting the most relevant features to generate interpretable Boolean feature vectors that explain the embeddings, allowing for explainable operations in technical systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are used to generate predictions for knowledge graph embeddings, then prediction accuracy is improved, but interpretability and explainability of the predictions deteriorate
Solution Approach 1:
The patent introduces an intermediary explanation generation module that translates the internal representations and decision processes of the machine learning model into human-interpretable explanations. This mediator component bridges the gap between the black-box predictions and human understanding, allowing users to comprehend the reasoning behind predictions without sacrificing prediction accuracy.
Solution Approach 2:
The patent segments the explanation generation process into distinct components: feature importance analysis, relationship path extraction, and explanation synthesis. By dividing the complex interpretation task into manageable segments, the system can provide detailed, step-by-step explanations that maintain both accuracy and interpretability.
2Adaptability or versatility
If complex machine learning models are deployed in technical systems, then system capabilities are improved, but the ability to detect and understand system behavior deteriorates
Solution Approach 1:
The patent implements feedback mechanisms that continuously monitor model predictions and generate explanatory feedback about the reasoning processes. This feedback loop allows operators to understand system behavior, verify predictions, and adjust parameters when necessary, thereby maintaining both advanced capabilities and detectability of system behavior.
Solution Approach 2:
The system performs self-explanation by automatically generating interpretations of its own predictions and decision-making processes. This self-service capability allows the complex model to communicate its own behavior and reasoning without requiring external analysis tools, making system behavior easily detectable and understandable.
Data Source
AI summary
A device and computer implemented method for machine learning. The method includes providing an embedding of an entity of a knowledge graph, determining a set of features for the entity depending on the knowledge graph, and providing the entity with a feature from the set of features that is selected depending on a score that is assigned to the feature with a model. The model is configured to map the set of features to a prediction for the embedding of the entity and to determine the score depending on a difference between the prediction and the embedding.