How AI Connects Different Databases Using Knowledge Graphs

This patent describes a server-based method that uses artificial intelligence and two learning models to automatically find and integrate connections between data fields and data values across multiple databases that have different structures.

What it covers

This patent describes a system for automatically integrating information from several databases, even if they are organized differently. First, the system creates 'knowledge graphs' for each database, which are like maps showing how data is structured (classes) and what specific data exists (instances) (Claim 1). These individual knowledge graphs are then fed into a 'first learning model' (an AI algorithm) to figure out how the data fields, or 'classes,' from different databases relate to each other. For example, it might learn that 'customer_ID' in one database is the same as 'client_number' in another. Next, the system uses a 'second learning model' to find connections between the actual data values, or 'instances,' across these databases, building on the class correlations already found (Claim 1). This results in a comprehensive, virtual integrated knowledge graph that can answer complex questions across all connected databases (Claim 9).

What it doesn't cover

• —Does not cover integrating databases without first generating knowledge graphs from them.
• —Does not cover systems that integrate databases using only one learning model to find both class and instance correlations simultaneously.
• —Does not cover manual methods of identifying correlations between data fields or values across databases.
• —Does not cover systems that rely solely on predefined schemas or mapping rules without using AI learning models to discover correlations.
• —Does not cover integrating databases where the learning models do not distinguish between correlations of 'classes' (data fields) and 'instances' (data values).

The clever bit

The clever part is using two distinct AI learning models: one specifically to find relationships between the *types* of data (classes) across different databases, and a second one to then find relationships between the *actual pieces of data* (instances), building on the first model's findings. This two-step, AI-driven approach automates a complex task that usually requires extensive manual effort.

Why it matters

In today's world, organizations often have many databases that don't talk to each other, making it hard to get a complete picture of information. This patent provides a way for AI to automatically find connections across these different data sources. This is crucial for businesses that need to combine customer data, sales figures, and inventory from various systems to make smarter decisions or offer better services. It helps overcome a major challenge in data management by creating a unified view.

Real-world examples

• 1.Enterprise data lakes and data warehouses
• 2.Customer Relationship Management (CRM) systems integrating with sales and support databases
• 3.Healthcare systems combining patient records from different clinics
• 4.Financial institutions linking transaction data from various departments
• 5.Supply chain management platforms integrating supplier and logistics databases