- Maintaining Common Semantics Across Sources
- 1 Purpose
- 2 Core Principles
- 3 Components of EA 2.0 Vocabulary System
- 4 Term Object Structure
- 5 Governance Workflow
- 6 Semantic Versioning
- 7 Synonym & Mapping Table
- 8 Integration with External Catalogs
- 9 Semantic Governance Metrics
- 10 Visualization
- 11 Common Pitfalls & Remedies
- 12 AI Augmentation
- 13 Benefits
- 14 Cultural Impact
- 15 Takeaway
Maintaining Common Semantics Across Sources #
1 Purpose #
Every enterprise speaks multiple dialects — business, technical, regulatory.
EA 2.0 acts as the interpreter, translating those dialects into a shared semantic model that all systems and teams can understand.
Vocabulary and taxonomy governance keep the meaning of entities — “Customer,” “Service,” “Incident,” “Control,” “Outcome” — consistent across cloud systems, BI dashboards, and AI reasoning layers.
Without this, AI can correlate data but never truly comprehend it.
2 Core Principles #
| Principle | Meaning |
|---|---|
| Single Semantic Backbone | One canonical term store feeding all domains. |
| Contextual Inheritance | Terms adapt per domain while preserving parent meaning. |
| Machine & Human Alignment | Taxonomies are readable by both business users and LLMs. |
| Change Control | Every definition change is versioned, approved, and propagated. |
| Traceable Usage | Every graph node references a term, making all data “typed.” |
3 Components of EA 2.0 Vocabulary System #
| Component | Description | Tooling / Source |
|---|---|---|
| Business Glossary | Defines business terms, KPIs, policies. | Collibra / SharePoint Term Store / Excel seed |
| Technical Catalog | Enumerates data entities, APIs, and schema names. | Purview / Snowflake / Data Factory |
| Risk & Control Taxonomy | Defines standardized risk types and mitigations. | ServiceNow GRC / ISO 27001 Mapping |
| Capability Taxonomy | Defines enterprise functions and outcomes. | EA 2.0 Ontology / Capability Map |
| Tag Dictionary | Maps informal labels or aliases to canonical terms. | EA 2.0 Graph extension |
Together, these components form EA 2.0’s semantic mesh.
4 Term Object Structure #
Each term in the EA 2.0 graph carries attributes:
| Field | Description | Example |
|---|---|---|
term_id | Unique key | TERM_12345 |
preferred_label | Canonical name | “Customer” |
definition | Official meaning | “Individual or organization purchasing goods/services.” |
aliases | Synonyms or variants | “Client,” “Buyer,” “Account” |
domain | Business / Technical scope | Sales |
version | Semantic version | 3.1 |
source_of_truth | Glossary source | Collibra |
last_reviewed | Date of stewardship approval | 2025-09-01 |
Every node (Capability, DataEntity, Policy, Risk, etc.) references at least one term_id.
5 Governance Workflow #
- Proposal: New term request submitted via form or Teams bot.
- Review: Steward checks duplication and domain alignment.
- Approval: Governance board validates definition.
- Propagation: EA 2.0 API syncs term to Graph, SharePoint, and Purview.
- Deprecation: Old terms flagged, relationships auto-repointed.
This workflow ensures controlled evolution rather than chaotic sprawl.
6 Semantic Versioning #
Each change increments term version (Major.Minor):
- Major (x.0) — definition meaning changed.
- Minor (.x) — formatting or metadata updated.
Relationships store which version they were created under, enabling semantic time-travel:
“Show all applications using the pre-2024 definition of ‘Customer.’”
7 Synonym & Mapping Table #
EA 2.0 maintains a Synonym Map:
| Alias | Canonical Term | Confidence | Source |
|---|---|---|---|
| Client | Customer | 0.95 | CRM API |
| Buyer | Customer | 0.9 | Procurement DB |
| Cust_ID | Customer | 0.8 | Legacy Schema |
Used by NLQ and RAG layers to interpret user prompts correctly — “show all clients” = Customer.
8 Integration with External Catalogs #
EA 2.0 exposes and consumes term metadata through APIs:
| System | Direction | Method | Purpose |
|:–|:–|:–|
| Collibra | Import / Sync | REST API / CSV export | Business glossary seed |
| Azure Purview | Bi-directional | Purview Lineage API | Data entity ↔ glossary link |
| SharePoint Term Store | Export | Graph API | Reuse in intranet / Teams |
| ServiceNow GRC | Import | Table API | Align risk/control taxonomies |
This federation keeps all sources semantically consistent without duplication.
9 Semantic Governance Metrics #
| KPI | Definition | Target |
|---|---|---|
| Term Coverage | % of nodes referencing valid terms | ≥ 95 % |
| Duplicate Terms | Terms with overlapping meanings | ≤ 2 % |
| Review Compliance | Terms reviewed in last 12 months | 100 % |
| Synonym Accuracy | Verified synonym-to-term mappings | ≥ 90 % |
| Cross-System Sync Lag | Time between term update and sync | < 24 h |
10 Visualization #
- Semantic Map: displays relationships among terms and synonyms.
- Domain Tree: hierarchical view of capabilities, risks, and data concepts.
- Change Timeline: shows evolution of term versions.
- Term Impact View: highlights all graph nodes affected by a definition change.
These are rendered in Power BI or directly in EA 2.0’s React front-end.
11 Common Pitfalls & Remedies #
| Issue | Effect | Remedy |
|---|---|---|
| Multiple glossaries per domain | Conflicting meanings | Central glossary + federation API |
| Missing term links | orphaned nodes | enforce term_id as mandatory field |
| Unapproved alias growth | inconsistent NLQ results | auto-detect aliases via LLM + steward validation |
| Over-engineered hierarchies | governance fatigue | keep ≤ 3 levels deep per domain |
12 AI Augmentation #
EA 2.0 uses AI to maintain its taxonomy intelligently:
- LLM detects potential duplicates or synonym clusters.
- NLP auto-suggests missing definitions or domain placements.
- Predictive tagging: new data feeds auto-linked to known terms with confidence scoring.
Human stewards approve AI-suggested terms — ensuring both speed and control.
13 Benefits #
✅ Consistent semantics across systems and analytics.
✅ Reliable NLQ responses (no synonym confusion).
✅ Easier integration with external data catalogs.
✅ Regulatory confidence through definition traceability.
✅ Faster onboarding for new architects and analysts.
14 Cultural Impact #
Vocabulary governance turns architecture from “diagramming” into shared literacy.
When everyone — from developer to CFO — means the same thing by service, collaboration accelerates.
15 Takeaway #
A shared language is the foundation of collective intelligence.
EA 2.0’s vocabulary governance makes the enterprise not only connected, but coherent — ensuring that every query, model, and decision speaks the same tongue.
