Knowledge Hypergraph Link Prediction via Virtual Node Segmentation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for link prediction in knowledge hypergraphs are not effective, as they are designed for knowledge graphs and fail to perform well when extended to hypergraphs, which store more complete information and are easier to obtain from relational databases.

Innovation Solution

A method and system for training an embedding model to perform link predictions in knowledge hypergraphs by introducing interactions between elements in the embedding space, enabling the model to represent relational algebra operations and improve link prediction performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If existing knowledge graph methods are extended to knowledge hypergraphs, then the methods can be applied to hypergraph data, but the link prediction performance deteriorates because these methods are not designed for hypergraph structures

Engineering Contradiction:

Improveapplicability to hypergraphsVSAvoidlink prediction performance

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent segments the hyperedge into individual binary relations by introducing virtual nodes, allowing existing knowledge graph methods to be applied while preserving hypergraph structure information. Each hyperedge (e1, e2, ..., en, h) is decomposed into multiple binary relations connecting entities to virtual nodes, enabling compatibility with standard KG embedding methods while maintaining hypergraph characteristics for accurate link prediction

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces virtual nodes as intermediaries to bridge the gap between hypergraph structures and traditional knowledge graph methods. These virtual nodes serve as mediators that connect multiple entities within a hyperedge, enabling the application of standard KG reasoning methods to hypergraph data while preserving the many-to-many relationship semantics of original hyperedges

Inventive Principle:

Principle #24Intermediary (Mediator)

2Loss of information

If all true information in the world is accessed and stored, then the knowledge hypergraph would be complete, but the difficulty of accessing and storing information increases

Engineering Contradiction:

Improvecompleteness of knowledgeVSAvoiddifficulty of accessing and storing

Core Design Contradiction:

Loss of informationVSDevice complexity

Solution Approach 1:

The patent creates a universal framework that handles both binary relations and non-binary relations through a unified hypergraph representation. The system can process various types of relationships (one-to-one, one-to-many, many-to-many) using the same hyperedge structure and virtual node mechanism, eliminating the need for separate storage and processing systems for different relation types

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Ease of operation

If knowledge hypergraphs are converted to knowledge graphs, then existing reasoning methods can be applied, but the performance of embedding-based methods deteriorates because they do not work well out of the box for converted hypergraphs

Engineering Contradiction:

Improveapplicability of existing methodsVSAvoidlink prediction accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent performs preliminary transformation of hypergraph data into a virtual node-based representation before applying embedding methods. By pre-processing the hypergraph structure to include virtual nodes and decomposed binary relations, the system prepares the data in a format that is compatible with existing KG embedding methods while preserving the information needed for accurate link prediction in the original hypergraph structure

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS12205002B2Method and system for training model to perform link prediction in knowledge hypergraph

Publication Date: 2025.01.21 SERVICENOW CANADA INC

AI summary

There is provided a method and system for training an embedding model to perform relation predictions in a knowledge hypergraph to output a trained embedding model. A training dataset comprising tuples representing relations between entities in the knowledge hypergraph are received. The embedding model is trained to perform relation predictions for each given tuple from a subset of tuples in the training dataset by generating a respective entity vector for each entity and a respective relation matrix representing relations between the entities. The entity vectors and relation matrix are split into a plurality of windows, and interaction values between elements in each window are calculated. A relation score indicative of the relation in the given tuple being true is calculated. Parameters of the embedding model are updated based on the relation scores for the subset of tuples. The trained embedding model is then output.