AI & Data Driven Enterprise
Collection of practical usage and demonstration heavy posts about the practical intersection of AI, Data, and Knowledge

DALL-E Generated

Reasoning & Inference — informed by Relationship Type Semantics

Created on 2025-09-27 04:02

Published on 2025-09-27 04:18

We hear a lot about symbolic reasoning and inference, but rarely see simple, live demonstrations of this fundamental and powerful concept.

Structured data, decomposed to its atomic parts, boils down to entity relationships. These relationships may exist persistently in some storage location (e.g., a DBMS or function-specific storage platform) or may be materialized at access time, with materialization itself depending on context. This could involve propositions expressed in a query language or a purpose-specific rules language.

For example, a SQL VIEW is a form of inference in a conventional RDBMS. Its limitations arise from the coarse-grained nature of the tuple structure (records in a table) around which entity relationships are organized.

RDF-based (or supporting) platforms offer the same functionality but leverage fine-grained entity relationships represented as entity → attribute → value or subject → predicate → object triples (3-tuples), split across ontology definitions and associated instances. This can be enhanced further by using hyperlinks to denote the represented entities and relationships.

With this foundation, we can return to reasoning and inference via simple examples, which was harder to demonstrate before LLMs emerged as powerful generic RDF clients. I share two important demos below.

Example 1 — British Royal Family Relationships

This example represents a snippet of known British Royal Family relationships, where the relationships are inferred—i.e., materialized at access time:

Reasoning enabled → https://linkeddata.uriburner.com/c/LkbC7nYdo

Here’s the same example with reasoning and inference disabled. In this case, you end up with a blank page until you re-enable the inference rule directive in the underlying query processor (currently commented out):

Reasoning disabled → https://openlinksw.com/c/5a5TSM8soK

The relevant pragma line is:

Remove the # (comment character) and rerun the query to see the inferred relationships via that link.
Why Is This Important?
In the age of AI and AI Agents, data is everything—the critical fuel that drives AI. For Agents in particular, the quality of their responses depends heavily on the contextual richness of the data they process.
See Also:
2016 post on LinkedIn regarding this demo → https://linkedin.com/pulse/reasoning-inference-using-british-royal-family-part-idehen/
Example 2 — Virtuoso & Neo4j Worldview Merging
Now that we have a basic example in place, here’s a more sophisticated demonstration showcasing the merging of worldviews across Virtuoso (a multi-model DBMS handling entity relationships as Tables or RDF-based Graphs) and Neo4j (a Labeled Property Graph DBMS).
In this case, Virtuoso connects to Neo4j and creates a Semantic Layer that harmonizes the external data, guided by an ontology. The end result? The ability to apply reasoning and inference to data residing in a system that doesn’t natively provide such capabilities.
Naturally, the same approach applies to:
Enabling Technologies from OpenLink Software
Related