ARI — Automated Relation Inference

Infer mappings between data shapes using semantic hints.

ARI (Automated Relation Inference) is a Cohesive product that proposes and ranks relationships between shapes—schemas, message structures, and domain models—by combining structural signals with a rich layer of semantics.

Instead of hand-authoring every field-to-field map, ARI uses your existing knowledge (code mappings, concept tags, semantic bindings, synonym expansion, and more) to infer candidate mappings with confidence and traceable rationale.

• ARI — Automated Relation Inference
• The problem: mapping is slow, brittle, and hard to maintain
• The solution: inference-driven mapping, grounded in your semantics
• Built on Cohesive.Relations, with an inference layer on top
• How ARI works
• What you get
• Flagship use case: inferring mappings from EDI to canonical domain models
• Where ARI fits
• Bring your semantics. Let ARI infer the relations.
• References

The problem: mapping is slow, brittle, and hard to maintain

Mapping between heterogeneous formats typically requires:

• Deep format expertise (e.g., EDI conventions vs. internal domain models)
• Extensive manual effort to find “equivalent” concepts that don’t share names
• Ongoing maintenance as partner specs and canonical models evolve
• Custom logic scattered across scripts, spreadsheets, and tribal knowledge

The result is predictable: long onboarding cycles, inconsistent transformations, and poor reuse.

The solution: inference-driven mapping, grounded in your semantics

ARI infers mappings between shapes by treating semantics as first-class signals.

You provide semantic hints such as:

• Code mappings: crosswalks between code sets, enumerations, qualifiers, and identifiers
• Concept tags: domain concepts attached to fields, segments, entities, and composites
• Semantic bindings: explicit links between local fields and shared concepts/ontologies
• Synonym expansion: controlled vocabulary, aliases, and terminology normalization
• Additional context: naming conventions, reference data, constraints, and structural patterns

ARI uses these hints to propose relationships between nodes in two shapes—then scores, explains, and iterates.

Built on Cohesive.Relations, with an inference layer on top

ARI uses Cohesive.Relations as its foundation for representing relationships between shapes. On top of that, ARI adds an inference layer powered by a Semantic Shape Graph:

• Start with a standard shape graph (structure: nodes, fields, hierarchy, cardinality)
• Attach semantic annotations to the same graph (concepts, bindings, tags, synonyms)
• Run inference to produce candidate relation edges between source and target shapes

This makes mapping a repeatable, inspectable process—not a one-off exercise.

How ARI works

1. Ingest shapes

Bring in source and target structures (e.g., EDI transaction structure and canonical domain model).

2. Attach semantics

Enrich shapes with concept tags, code crosswalks, bindings, and synonym expansions.

3. Infer candidate relations

ARI proposes mappings by combining:

1. Structural alignment (shape/position/context)
2. Semantic alignment (concept overlap, binding compatibility, synonym proximity)
3. Value-level cues (code sets, qualifiers, identifiers)
4. Score + explain

Each candidate mapping includes confidence and supporting evidence (which hints contributed and why).

5. Refine and operationalize

Confirm, override, or constrain results; export relations for downstream transformation and governance.

What you get

• Candidate mappings between shapes (field, element, and concept-level)
• Confidence scoring for prioritization and review workflows
• Explainability tied to semantic hints (tags, bindings, synonyms, code crosswalks)
• Consistency across partners and message versions by reusing semantic assets
• Change resilience when either side evolves (re-run inference with preserved semantics)

Flagship use case: inferring mappings from EDI to canonical domain models

ARI’s flagship deployment targets one of the highest-friction mapping problems: EDI → canonical domain model.

EDI mapping is hard because equivalence is rarely name-based. Meaning is spread across:

• Qualifiers and codes
• Segment context
• Loops and repeated structures
• Partner-specific conventions

ARI addresses this by grounding inference in semantic signals (especially code mappings and concept tags), enabling it to:

• Align EDI elements to canonical fields even when names diverge
• Reconcile qualifier-driven meaning with canonical intent
• Accelerate partner onboarding by reusing semantic hints across transactions
• Keep canonical models stable while supporting many EDI variants

Outcome: reduced manual mapping effort, faster onboarding, and more reusable integration logic.

Where ARI fits

ARI is designed for teams building and maintaining data interchange at scale:

• EDI modernization programs
• Canonical model standardization initiatives
• Integration platforms and connector factories
• Data product and governance teams that need traceability

While EDI is the flagship, the same approach applies to APIs, events, files, and internal schemas anywhere “shape-to-shape” mapping is required.

Bring your semantics. Let ARI infer the relations.

If you already have code crosswalks, glossaries, tags, or domain bindings—ARI converts that knowledge into inferred mappings you can review, govern, and reuse.

References

Cohesive Relations

Cohesive Building Blocks