Knowledge Graph Data Integration for Relational Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In relational database management systems, data analysis becomes unwieldy due to disparate data formats, duplicate entries, and lack of integration, leading to resource consumption, inaccurate analytics, and difficulty in searching and understanding relevant data across different sources.
Innovation Solution
A central data model that integrates physical data elements using a semantic model, enabling users to search and explore data relationships through a knowledge graph server with machine learning models for data translation, deduplication, and semantic searching, which harmonizes data formats and links disparate data sources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If data is stored in multiple relational tables with different formats and column labels, then data can be collected from various sources, but data analysis becomes unwieldy and resource consumption increases
Solution Approach 1:
The patent merges multiple disparate data sources and formats into a unified knowledge graph structure. Different tables with varying column labels (e.g., 'birthdate', 'b-date') are integrated into a single standardized model, allowing efficient querying and analysis without the overhead of managing multiple separate data sources.
Solution Approach 2:
The knowledge graph serves as an intermediary layer between raw relational data and analysis tools. It translates and harmonizes data from various sources into a unified semantic model, enabling efficient data retrieval and analysis while preserving the adaptability to collect data from diverse origins.
2Adaptability or versatility
If data is stored in multiple tables with different column labels and formats, then data from various sources can be integrated, but searching and understanding relevant data becomes difficult
Solution Approach 1:
The patent applies homogeneity by standardizing disparate data formats and column labels into a unified knowledge graph schema. All birthdate variations (dd-mm-yyyy, mm-dd-yyyy, yyyy-mm-dd) are normalized to a single format, and all column labels are mapped to consistent semantic entities, making data searching and understanding straightforward.
Solution Approach 2:
The knowledge graph provides a universal interface for accessing data from multiple sources. It offers a single standardized query language and data model that works across all integrated data sources, eliminating the need for source-specific search procedures and improving ease of operation.
3Quantity of substance
If data is collected from many different tables and formats, then comprehensive data coverage is achieved, but duplicate entries and inaccurate analytics increase
Solution Approach 1:
The patent extracts and identifies duplicate entries during the knowledge graph construction process. By comparing standardized data elements across multiple sources, it detects and removes redundant records, ensuring that comprehensive data coverage is maintained while eliminating duplicates that would compromise analytics accuracy.
Solution Approach 2:
The system implements feedback mechanisms to verify data quality during integration. As data from multiple sources is incorporated, the knowledge graph continuously validates entries against existing records and semantic constraints, providing feedback to identify and correct duplicates or inconsistencies, thereby maintaining high data accuracy.
Data Source
AI summary
A method may include retrieving, over a network device, data from a data source, the data formatted according to a relational database schema; converting, using at least one processor, the data from the data source into a knowledge graph; training, using the at least one processor, a neural network using triples from the knowledge graph, wherein a loss function of the neural network is based on a distance between a true triple from the knowledge graph and a corrupt triple; and after the training, storing weights of a hidden layer of the neural network as a vector space representation of entities in the knowledge graph.