Technology

We build a high-level knowledge model that defines concepts and relationships across structured and unstructured data. By reconciling different names and codes for the same entity across systems, we establish consistent interpretation.

We use an ontology to connect real data as nodes and edges, creating a knowledge network that makes hidden relationships and context available for exploration.

We identify patterns through graph neural networks (GNNs) and rule-based reasoning. GraphRAG connects LLM responses to a knowledge graph to reduce hallucinations and provide supporting evidence.

Using PR-OWL and Multi-Entity Bayesian Networks (MEBN), we represent probability distributions and causal relationships within ontologies. We combine knowledge fragments relevant to a query and update inferences as new evidence arrives, providing probability estimates together with their basis.