Relational Database Embedding via Entity-Relation Matrices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods require the generation of pseudo-documents or graphs on top of relational databases before embedding learning can be used to extract information, which is inefficient and cumbersome.
Innovation Solution
A system and method that directly generates entity-relation matrices from relational databases and performs embedding learning processes on these matrices to update embeddings associated with entities, eliminating the need for intermediate steps like pseudo-document or graph generation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pseudo-documents or graphs are generated on top of relational databases before embedding learning, then embedding learning can extract information from the database, but the process becomes inefficient and cumbersome
Solution Approach 1:
The patent extracts the essential relational structure directly from the database into entity-relation matrices, eliminating the need for intermediate pseudo-document or graph generation steps. This direct extraction approach maintains information extraction capability while removing unnecessary preprocessing overhead.
Solution Approach 2:
Instead of converting relational data into pseudo-documents or graphs before embedding learning, the patent inverts the approach by directly applying embedding learning to entity-relation matrices derived from the relational database schema, thus streamlining the workflow.
2Reliability
If pseudo-documents or graphs are generated as intermediate steps, then embedding learning can be applied, but the process becomes cumbersome
Solution Approach 1:
The patent extracts relational data directly into entity-relation matrices that are suitable for embedding learning, removing the need for complex intermediate conversions to pseudo-documents or graphs. This reduces process complexity while maintaining embedding learning applicability.
Solution Approach 2:
The patent changes the parameter representation by directly using entity-relation matrices with numerical attributes suitable for embedding learning, eliminating the need for complex data structure transformations and reducing overall process complexity.
3Productivity
If direct embedding learning is performed on relational databases without preprocessing, then efficiency is improved, but the method must handle raw relational data directly
Solution Approach 1:
The patent segments the relational database into entity-relation matrices that separate entities, relations, and attributes into structured components suitable for embedding learning. This segmentation simplifies the implementation by organizing raw relational data into a format that is directly compatible with embedding algorithms.
Solution Approach 2:
The patent transforms relational data parameters into a format suitable for embedding learning by creating entity-relation matrices with appropriate numerical representations, making the implementation straightforward while maintaining high processing efficiency.
Data Source
AI summary
Provided are systems for analyzing a relational database using embedding learning that may include at least one processor programmed or configured to generate one or more entity-relation matrices from a relational database and perform, for each entity-relation matrix of the one or more entity-relation matrices, an embedding learning process on an embedding associated with an entity. When performing the embedding learning process on the embedding associated with the entity, the at least one processor is programmed or configured to generate an updated embedding associated with the entity. Computer-implemented methods and computer program products are also provided.