How to Combine Information from Multiple Knowledge Graphs

This patent describes a method for merging information from two or more existing knowledge graphs into a single, unified knowledge graph, using special rules called "graph operators."

What it covers

The patent outlines a method for combining information from different knowledge graphs. First, it obtains a "target entity field" (like "person" or "product") and a "target relationship description" (like "works for" or "buys") from the "ontology definition data" of two or more existing knowledge graphs (Claim 1). Then, it determines "graph operators" which are specific rules for combining these fields and relationships (Claim 1). It gathers "data instances" (actual data, like "John Doe" or "iPhone") corresponding to these fields and relationships from the original graphs (Claim 1). Finally, it uses the graph operators to process these data instances and create a "fused knowledge graph" (Claim 1). For example, if one graph has "customer" and another has "user," a graph operator could fuse them into a single "customer/user" entity, standardizing how their attributes are expressed (Claim 5).

What it doesn't cover

• —Does not cover fusing data that is not structured as a knowledge graph with defined entities and relationships.
• —Does not cover data fusion methods that do not use specific "graph operators" to process data instances.
• —Does not cover combining knowledge graphs without first selecting target entity fields and relationship descriptions from their ontology definitions.
• —Does not cover simple data aggregation without creating a new, fused knowledge graph structure.
• —Does not cover fusion where the operators are not determined either by user input or automatic generation (Claim 3).

The clever bit

The novelty lies in systematically using "graph operators" to not just combine raw data, but to specifically fuse the *ontology definition data* (the blueprints) and then process the *data instances* (the actual facts) from multiple knowledge graphs, ensuring a coherent and structured merged result.

Why it matters

In today's world, information is often scattered across many different databases and systems. This patent provides a structured way to bring together knowledge from various sources, like different departments in a company or various online platforms. This allows for a more complete understanding of complex topics, making it easier to find connections and draw conclusions that wouldn't be possible with isolated data.

Real-world examples

• 1.Combining customer data from an e-commerce platform and a payment system.
• 2.Merging product catalogs from different suppliers in a supply chain.
• 3.Integrating medical research data from various hospital systems.
• 4.Creating a unified view of financial transactions from multiple banking services.