I have an existing knowledge graph in Neo4j where each node (e.g., Person, Organization) is uniquely identified by a UUID. I’m planning to use the Neo4j KGB (Knowledge Graph Builder) to ingest new content into the graph.
Here’s what I’m trying to achieve:
-
As the KGB processes and chunks documents, if a named entity extracted from the chunk already exists in my graph (based on UUID or name), I want the new chunk to be associated with that existing entity node.
-
I want to prevent KGB from creating a duplicate node for an entity that already exists.
My Questions:
-
How can I prepare my data to ensure that chunked content is linked to pre-existing nodes?
-
Should I enrich the
entities.csvfile with UUIDs of my existing nodes and ensure KGB recognizes them? -
Does KGB support an
idorexternalIdmapping in its import process to resolve against existing nodes?
-
-
Is there a way to customize the entity matching logic (e.g., match by UUID or name similarity) during the
kgb importstage? -
If I have pre-chunked content, is there a way to feed that into the pipeline and explicitly control the node associations?
Thanks in advance for your help.