Managed Semantic Objects for Knowledge Graph Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current knowledge graphs (KGs) and ontologies face challenges in representing real-world relationships effectively, lacking accuracy and detail, and are limited in generating sufficient training parameters for AI and ML models, particularly in capturing domain-specific knowledge and interdependencies, which hinders their ability to generalize to diverse enterprise problems.
Innovation Solution
The implementation of managed semantic objects (MSOs) within a knowledge graph (KG) that include containers expressing features of real-world objects, enhanced with metadata and tags to represent relationships, evidence, and inference, enabling the creation of more accurate and actionable training parameters by incorporating tribal and institutional knowledge.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional knowledge graphs are used to represent real-world relationships, then the system structure is simple, but the accuracy and detail of relationship representation is insufficient
Solution Approach 1:
The patent segments the knowledge representation into multiple hierarchical levels: entities, attributes, relationships, and contextual metadata. Each relationship is broken down into source entity, target entity, relationship type, evidence, and inference components. This segmentation enables precise representation of real-world relationships while maintaining manageable system structure through modular organization.
Solution Approach 2:
The patent adds dimensional depth to traditional knowledge graphs by introducing contextual dimensions such as evidence sources, inference confidence levels, temporal validity, and spatial context. This transforms flat relationship representations into multi-dimensional structures that capture nuanced real-world relationships with greater accuracy without proportionally increasing complexity.
2Reliability
If more domain-specific knowledge and interdependencies are incorporated, then the training parameters for AI and ML models improve, but the processing complexity and resource demands increase
Solution Approach 1:
The patent performs preliminary organization and structuring of domain-specific knowledge during the knowledge graph construction phase. Relationships are pre-tagged with metadata including evidence sources, confidence levels, and contextual attributes. This preliminary action enables AI and ML models to access well-structured training parameters without requiring complex processing during inference, thus improving model reliability while managing processing complexity.
Solution Approach 2:
The patent applies different levels of detail and processing to different parts of the knowledge graph based on local requirements. High-priority relationships with critical domain knowledge receive enhanced metadata and verification, while less critical relationships use simplified representations. This local quality approach ensures that processing resources are focused where they most improve AI/ML model performance without uniformly increasing complexity across the entire system.
3Adaptability or versatility
If diverse knowledge sources including tribal and institutional knowledge are integrated, then the quantity and diversity of training parameters increase, but the time and resources required for integration increase
Solution Approach 1:
The patent implements a universal metadata schema and standardized relationship representation that can accommodate diverse knowledge sources including tribal knowledge, institutional knowledge, scientific literature, and expert opinions. This multi-functional framework enables integration of various knowledge types without requiring separate processing pipelines for each source, thus increasing knowledge diversity while reducing integration time through standardized procedures.
Solution Approach 2:
The patent creates structured representations (copies) of knowledge from diverse sources that preserve the essential characteristics and relationships while fitting into a unified knowledge graph framework. Instead of integrating raw diverse data formats directly, the system creates standardized copies with consistent metadata structures, enabling efficient integration of tribal and institutional knowledge without proportionally increasing integration time.
Data Source
AI summary
Briefly, embodiments, such as methods and/or systems for creating and/or updating elements of a knowledge graph (KG), for example, are described.