Knowledge Graph Reasoning via Type Distribution Screening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Knowledge graph reasoning efficiency is low due to the large number of entities in knowledge graphs, leading to inefficient traversal and incomplete knowledge graphs that may miss potential entity relationships.
Innovation Solution
A method that selects nearest neighbor entities based on proximity and type distribution probabilities, using a type distribution prediction model to optimize model parameters and reduce the number of candidate entities, thereby improving reasoning efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all entities in the knowledge graph are traversed to perform reasoning, then the completeness of reasoning results is improved, but the reasoning efficiency deteriorates due to the large number of entities
Solution Approach 1:
The patent segments the large set of all entities into smaller subsets based on type distribution. It divides entities into different type categories (e.g., person, organization, location) and processes each type separately, allowing parallel computation and reducing the traversal burden while maintaining completeness within each type segment
Solution Approach 2:
The patent performs preliminary classification of entities by type before the actual reasoning traversal. By pre-grouping entities according to their types and predicting type distributions, the system prepares the data structure in advance to enable more efficient targeted traversal rather than exhaustive search across all entities
2Productivity
If the number of candidate entities is reduced to improve reasoning efficiency, then the reasoning speed is improved, but the accuracy of finding correct relationships deteriorates
Solution Approach 1:
The patent applies different processing strategies to different types of entities based on their local characteristics. Each entity type (person, organization, location, etc.) is handled with type-specific reasoning rules and probability thresholds, allowing the system to maintain high accuracy for each category while overall reducing the candidate set size
Solution Approach 2:
The patent changes the parameter of entity selection from binary inclusion/exclusion to probability-based filtering. By calculating type distribution probabilities and using threshold-based filtering, the system dynamically adjusts which entities become candidates, maintaining accuracy through probabilistic reasoning while reducing the absolute number of candidates processed
3Device complexity
If traditional reasoning methods are used without type distribution prediction, then the implementation complexity is reduced, but the ability to handle incomplete knowledge graphs deteriorates
Solution Approach 1:
The patent implements self-service through automated type distribution prediction that requires no manual labeling or external annotations. The system automatically learns entity type distributions from the knowledge graph structure itself and uses this self-derived information to guide reasoning, making the system self-sufficient while improving robustness to incompleteness
Solution Approach 2:
The patent incorporates feedback mechanisms where the predicted type distribution information is continuously refined based on reasoning results. The system uses feedback from the reasoning process to adjust and improve its type distribution predictions, creating a self-improving loop that enhances reliability without proportionally increasing complexity
Data Source
AI summary
Implementations of the present specification disclose a knowledge graph reasoning method and apparatus, a model training method and apparatus, and a computer device. The method includes: obtaining a query entity and a query relationship; selecting one or more nearest neighbor entities of the query entity from a knowledge graph; determining a first probability of a nearest neighbor entity of the one or more nearest neighbor entities, where the first probability is used to indicate a possibility that the nearest neighbor entity is in communication with the query relationship; selecting a nearest neighbor entity of the one or more nearest neighbor entities as a candidate entity based on the first probability; and selecting a candidate entity matching the query entity and the query relationship as a result entity. In the implementations of the present specification, the efficiency of knowledge graph reasoning can be improved.