Clustering Inference Rules for Knowledge Graph Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for grouping entities into clusters in knowledge graphs face challenges in ensuring the quality and explainability of the grouping, particularly for diffuse clusters and incomplete knowledge graphs, where predefined clusters are difficult to describe effectively.
Innovation Solution
A computer-implemented method that iteratively determines associations between target entities and clusters using inference rules, with a feedback mechanism to refine associations based on a threshold value, and incorporates a multidimensional state space mapping to improve clustering precision and quality, allowing for the selection of high-quality inference rules and symbolic descriptions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If deep neural networks are used to group data in high-dimensional state spaces, then clustering capability is improved, but the quality and explainability of the grouping deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where the determined cluster associations are fed back into the system to refine and improve subsequent clustering results. The system evaluates the quality of associations and uses this information to adjust the clustering process iteratively, thereby maintaining high clustering capability while improving grouping quality and explainability through continuous refinement based on feedback from association quality assessment.
2Ease of manufacture
If clusters are determined based on statistical testing, then clustering process is simplified, but the precision of cluster-description association deteriorates
Solution Approach 1:
The patent performs preliminary extraction of patterns that define clusters before determining associations between clusters and descriptions. By pre-processing the data to identify meaningful patterns and structures, the system simplifies the subsequent association determination while ensuring high precision in matching clusters with appropriate descriptions, as the foundational pattern extraction is already optimized for accuracy.
3Productivity
If predefined clusters are used, then clustering efficiency is improved, but the explainability of diffuse clusters deteriorates
Solution Approach 1:
The patent introduces a new dimension of pattern-based representation for clusters. Instead of relying solely on predefined cluster labels, the system extracts and utilizes patterns that define each cluster, adding a descriptive dimension that enhances explainability. This pattern dimension allows diffuse clusters to be explained through their defining characteristics rather than just their predefined labels, maintaining efficiency while recovering lost information.
4Measurement precision
If multiple evaluation metrics are used to select patterns, then description accuracy is improved, but the complexity of the procedure increases
Solution Approach 1:
The patent segments the complex procedure into distinct modules: pattern extraction, pattern evaluation using multiple metrics, pattern selection, and association determination. By dividing the overall process into these manageable segments, the system can employ multiple evaluation metrics for accurate description selection without overwhelming complexity, as each segment handles a specific aspect of the task with well-defined inputs and outputs.
Data Source
AI summary
A computer-implemented method for grouping target entities into clusters. A base association in which a cluster is associated with each of the target entities is determined in a computation step for the target entities as a function of an association for entities. Inference rules are determined as a function of the association for entities and as a function of the base association, each of the inference rules defining an association of entities with one of the clusters. An altered association is determined as a function of the association for entities and the inference rules. A check is made as to whether a difference between the base association and the altered association falls below a threshold value. When it does, an association of the target entities with the clusters is output or stored. Otherwise, a feedback value is determined as a function of the difference.