How Find Connections Works

The service follows a multi-phase pipeline:

1. Discovery — Enumerates all classes and their datatype properties within the graphmart, collecting property-level statistics (total values, unique values).
2. Pivot Layer Creation — Creates a temporary layer in the graphmart that materializes property values into a normalized structure optimized for cross-class comparison.
3. Pair Matching — For each pair of classes, compares their property values using the selected matching strategy to find the best-matching property pair.
4. Best Match Selection — Selects the property pair with the highest combined source and target overlap that meets the configured minimum overlap thresholds.
5. Ontology and Step Creation — Creates a new link ontology defining the discovered object properties and generates SPARQL INSERT transform steps in a new graphmart layer.
6. Cleanup — Removes the temporary pivot layer and refreshes the graphmart.