Knowledge Graph Embedding Prior Knowledge Extraction Framework
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing knowledge graph embedding technologies face limitations in improving performance across various models, requiring code corrections and often breaking the inherent properties of existing models during ensemble methods.
Innovation Solution
A framework system that includes a basic learning unit, prior knowledge extraction unit, and enhanced learning unit to improve knowledge graph embedding models by initializing or transforming embedding vectors based on prior knowledge extracted from input data, allowing for performance enhancement without altering the existing model's properties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing knowledge graph embedding models are improved through ensemble methods or code corrections, then performance improvement is achieved, but the inherent properties of the models are broken and implementation becomes complex
Solution Approach 1:
The patent applies preliminary action by extracting prior knowledge from the knowledge graph before the main embedding learning process. This prior knowledge extraction prepares the data in advance to guide the embedding model, improving performance without requiring complex modifications to the model structure or ensemble methods during the main processing stage.
Solution Approach 2:
The patent introduces an intermediary component - the prior knowledge extraction module - that acts as a mediator between the raw knowledge graph data and the embedding model. This intermediary extracts and processes prior knowledge separately, then feeds it to the embedding model, avoiding direct modification of the model's inherent properties while still achieving performance improvement.
2Measurement precision
If knowledge graph embedding models are improved through model-specific methods, then performance improvement is achieved, but adaptability to different models is reduced
Solution Approach 1:
The patent applies universality by designing a prior knowledge extraction framework that is model-agnostic and can be applied to various knowledge graph embedding models. The extraction process focuses on general knowledge graph structures and relationships rather than model-specific features, making the improvement method universally applicable across different embedding models without requiring model-specific customization.
3Measurement precision
If prior knowledge extraction is performed from embedding vectors, then performance improvement is achieved, but additional processing time is required
Solution Approach 1:
The patent performs prior knowledge extraction as a preliminary step before the main embedding learning process. By extracting and processing prior knowledge in advance, the system prepares optimized input data that accelerates the subsequent embedding learning, reducing the overall training time despite the additional initial processing.
Solution Approach 2:
The system performs self-service by automatically extracting prior knowledge from the knowledge graph structure itself without requiring external annotations or manual intervention. This automated approach minimizes additional time investment while maximizing the benefit of prior knowledge utilization.
Data Source
AI summary
A learning method for improving performance of a knowledge graph embedding model is provided. The method includes: performing learning of a first knowledge graph embedding model based on input knowledge data; extracting all embedding vectors from the learned first knowledge graph embedding model, and extracting prior knowledge based on the extracted embedding vectors; and performing learning of a second knowledge graph embedding model through at least one of initialization of the embedding vectors and transform of the input knowledge data based on the extracted prior knowledge.