This is the multi-page printable view of this section. Click here to print.

Return to the regular view of this page.

Ontological Semantics

A theory of meaning in natural language and an approach to NLP that uses a constructed world model — an ontology — as a central resource for extracting, representing, and reasoning about knowledge derived from natural language texts.

1: Paper on Glass: Digital Process Data Capture with S88/S95 Ontology Alignment

Architecture of Ontological Semantics#

The key components required to build a knowledge framework leveraging Ontological Semantics — the theoretical grounding that has enabled the current field of Large Language Models — which help us create knowledge out of existing sources in a systematic manner.

@startuml
!theme plain
skinparam defaultFontSize 13
skinparam component {
  BackgroundColor #EBF5FB
  BorderColor #2980B9
  FontColor #1A252F
  RoundCorner 10
}
skinparam database {
  BackgroundColor #FEF9E7
  BorderColor #F39C12
}
skinparam rectangle {
  BackgroundColor #EAFAF1
  BorderColor #27AE60
  RoundCorner 8
}
skinparam arrow {
  Color #2C3E50
  FontColor #555
  FontSize 11
}

title Architecture of Ontological Semantics

package "Static Knowledge Sources" {
  database "Ontology\n(World Model)" as ONT
  database "Fact Repository\n(Domain Facts)" as FACT
  database "Lexicon\n(Word Forms & Meanings)" as LEX
  database "Onomasticon\n(Named Entities)" as ONOM
}

component "Knowledge Representation\nLanguage" as KRL

rectangle "Semantic Analyzer\n(Large Language Model)" as LLM

rectangle "Use Cases" as UC {
  component "Text Summarization" as TS
  component "Question & Answering" as QA
  component "AI Agents" as AG
}

ONT --> KRL : "defines concepts"
FACT --> KRL : "instantiates"
LEX --> KRL : "maps words → meaning"
ONOM --> KRL : "resolves named entities"

KRL --> LLM : "structured meaning\nrepresentation"

LLM --> TS
LLM --> QA
LLM --> AG
@enduml

The Four Static Knowledge Sources#

Ontology — the World Model#

The ontology defines the concepts that exist in a domain and the relationships between them. It is the schema of the knowledge graph: what nodes (concepts) and edge types (relationships) are valid. Without an ontology, a knowledge base is just a bag of facts with no shared meaning.

Fact Repository#

Stores instantiated knowledge — specific, asserted facts about the world derived from texts, experiments, or expert input. Facts are the populated rows of the ontology’s schema.

Lexicon#

Maps surface word forms to their semantic meanings. A lexicon allows the system to recognize that “kinase inhibitor,” “enzyme blocker,” and “phosphorylation suppressor” may all refer to the same ontological concept, enabling robust natural language understanding across varied terminology.

Onomasticon#

A specialized lexicon for named entities: people, organizations, places, products, and standards (e.g., “ICH Q10,” “S88,” “FDA”). The onomasticon lets the system resolve ambiguous proper nouns to canonical knowledge graph nodes.

Knowledge Representation Language#

The Knowledge Representation Language (KRL) is the formal grammar that ties the four sources together. It allows a system to derive meaning from raw text by:

Parsing surface language through the Lexicon
Resolving named entities through the Onomasticon
Grounding concepts against the Ontology
Asserting new facts into the Fact Repository

This pipeline is what transforms unstructured text into structured, queryable knowledge graph entries.

Semantic Analyzer — Large Language Models#

The Semantic Analyzer uses the KRL-structured knowledge to drive user-facing interactions. Modern LLMs play this role, but a key distinction applies:

@startuml
!theme plain
skinparam defaultFontSize 13
skinparam rectangle {
  RoundCorner 10
  FontColor white
}

rectangle "**Black Box LLM**\n(no visible ontology)\n\nGenerates answers\nCannot explain reasoning\nKnowledge is opaque" as BB #C0392B
rectangle "**Glass Box (Ontological Semantics)**\n(explicit ontology + fact repository)\n\nGenerates answers\nCan cite sources\nKnowledge is auditable" as GB #27AE60

BB -right[hidden]-> GB
note bottom of BB : Hallucination risk\nLow traceability
note bottom of GB : Ontology-grounded\nHigh explainability
@enduml

The Glass Box advantage is critical in regulated domains like pharma, manufacturing, and clinical research — where every knowledge claim must be traceable to its source.

Supported Use Cases#

Use Case	How Ontological Semantics Helps
Text Summarization	Extracts key ontology nodes and relationships from documents, producing summaries that preserve semantic fidelity
Question & Answering	Grounds answers in the Fact Repository and Ontology rather than statistical patterns alone
AI Agents	Agents navigate the knowledge graph to plan multi-step reasoning, with each hop traceable to an ontology edge
Ontology Alignment	Maps domain-specific standards (S88, S95, ICH) to a shared ontology for interoperability

Applied Example: This Site’s Taxonomy as an Ontology#

The STEAM knowledge graph on this site is itself an instance of ontological semantics in practice:

Ontology nodes: Science, Technology, Engineering, Arts, Mathematics
Ontology edges: enables, implements, models, formalized by, grounded in
Lexicon: tags and categories on each page
Onomasticon: persona names, standard names (ICH, S88, S95)
Fact Repository: each page’s related_concepts frontmatter

See the full graph: STEAM Knowledge Graph →

1 - Paper on Glass: Digital Process Data Capture with S88/S95 Ontology Alignment

How S88 batch control and S95 enterprise-integration ontologies align with paper-on-glass interfaces to create a unified, semantically rich manufacturing knowledge framework.

Abstract#

The “paper on glass” paradigm represents a transformative approach to industrial process documentation, where traditional paper-based records are digitally replicated while maintaining familiar workflows. This paper explores how such systems can be enhanced through alignment with established manufacturing ontologies, specifically ISA-88 (S88) batch control and ISA-95 (S95) enterprise-control integration standards, creating a unified framework for recipe management and process data capture.

Introduction#

Manufacturing industries have long relied on paper-based documentation for recipe management, batch records, and process data collection. The “paper on glass” concept digitizes these familiar interfaces while preserving the cognitive patterns operators have developed over decades. However, true digital transformation requires more than visual replication—it demands semantic alignment with standardized ontologies that enable interoperability, data consistency, and automated processing.

The Paper on Glass Experience#

Visual Familiarity with Digital Power#

The paper on glass interface maintains the visual metaphors of traditional documentation:

Form-based layouts that mirror paper worksheets
Signature fields and approval workflows
Checkbox lists and manual data entry points
Familiar navigation patterns and information hierarchy

However, beneath this familiar surface lies a sophisticated data model that captures not just values, but semantic meaning, relationships, and process context.

Contextual Data Capture#

Unlike simple form digitization, true paper on glass systems capture:

Temporal context: When actions occurred relative to process phases
Causal relationships: How parameters influence outcomes
Operational context: Who performed actions and under what conditions
Equipment state: The configuration and status of process equipment
Material genealogy: Traceability of inputs through transformation

S88 Ontology Integration#

Recipe Hierarchy Alignment#

The S88 standard defines a clear hierarchy for batch processes:

Recipe Level → Procedure → Unit Operation → Operation → Phase

Paper on glass systems aligned with S88 map traditional recipe cards to this structure:

Master Recipe: The idealized process definition
Control Recipe: Site-specific adaptation with equipment bindings
Batch Record: Execution instance with actual parameters and results

Procedural Control Elements#

S88’s procedural control elements provide semantic structure for recipe steps:

Phases become interactive workflow steps in the digital interface
Operations group related phases with clear start/end conditions
Unit Operations align with equipment modules and process cells
Procedures represent complete processing sequences

Equipment Entity Modeling#

The S88 equipment model maps directly to paper on glass interfaces:

Process Cells become top-level organizational units
Units correspond to major equipment systems
Equipment Modules represent controllable subsystems
Control Modules map to individual instruments and actuators

S95 Enterprise Integration#

Functional Hierarchy#

S95 defines four levels of manufacturing operations:

Level 4: Business planning and logistics
Level 3: Manufacturing operations management
Level 2: Supervisory control
Level 1: Basic control

Paper on glass systems typically operate at Levels 2-3, bridging operator interfaces with enterprise systems.

Information Models#

S95 information models provide structure for:

Work Orders: Linking recipes to production schedules
Material Definitions: Standardizing ingredient and product specifications
Equipment Information: Maintaining asset hierarchies and capabilities
Production Performance: Capturing efficiency and quality metrics

Activity Models#

S95 activity models define standard operations:

Production Scheduling: Translating demand into executable recipes
Production Dispatching: Assigning resources to batch executions
Production Execution: Real-time monitoring and control
Production Tracking: Historical analysis and compliance reporting

Unified Recipe Management Framework#

Recipe as Executable Specification#

When aligned with S88/S95 ontologies, recipes become more than instructions—they become executable specifications that:

Define precise control sequences
Specify material requirements and constraints
Establish quality checkpoints and acceptance criteria
Enable automatic equipment configuration
Support predictive quality modeling

Semantic Data Binding#

The ontological alignment enables semantic data binding where:

Process parameters are linked to their physical measurements
Material additions are tracked through inventory systems
Quality attributes are connected to analytical results
Deviations trigger defined response procedures
Historical data supports continuous improvement

Cross-Recipe Learning#

Standardized ontologies enable learning across recipes:

Common phases can share optimization insights
Material properties influence multiple formulations
Equipment performance patterns apply across products
Quality correlations span recipe families

Implementation Architecture#

Ontology Layer#

The foundation consists of:

S88 Process Model: Defining procedural hierarchy and equipment entities
S95 Information Model: Structuring materials, personnel, and equipment data
Domain Extensions: Industry-specific concepts and relationships
Semantic Mappings: Linking legacy data to standardized concepts

Data Capture Layer#

This layer provides:

Contextual Forms: Paper-like interfaces with semantic data binding
Workflow Engine: S88-compliant procedure execution
Real-time Integration: Connecting to process control systems
Validation Rules: Ensuring data quality and compliance

Analytics Layer#

Advanced capabilities include:

Process Mining: Discovering optimization opportunities from execution data
Predictive Quality: Using historical patterns to predict outcomes
Anomaly Detection: Identifying deviations from normal operation
Continuous Improvement: Systematic recipe refinement

Benefits and Outcomes#

Operational Excellence#

Reduced Training Time: Familiar interfaces with enhanced capabilities
Improved Compliance: Automated validation and audit trails
Faster Problem Resolution: Structured data enables root cause analysis
Enhanced Flexibility: Rapid recipe modification and deployment

Strategic Advantages#

Data Standardization: Consistent semantics across operations
System Interoperability: Standards-based integration
Knowledge Preservation: Structured capture of process expertise
Innovation Acceleration: Data-driven recipe development

Challenges and Considerations#

Change Management#

Balancing familiar interfaces with new capabilities
Training operators on enhanced functionality
Managing the transition from paper-based workflows
Ensuring buy-in from production teams

Technical Implementation#

Integrating with existing control systems
Ensuring real-time performance requirements
Managing data volume and storage requirements
Maintaining system reliability and availability

Organizational Alignment#

Coordinating IT and OT system integration
Establishing data governance frameworks
Managing intellectual property and security concerns
Aligning with regulatory requirements

Future Directions#

AI-Enhanced Recipe Development#

Machine learning algorithms operating on S88/S95-structured data can:

Automatically optimize process parameters
Predict quality outcomes from input conditions
Suggest recipe modifications based on historical performance
Identify opportunities for process innovation

Digital Twin Integration#

Paper on glass interfaces can serve as control points for digital twins:

Real-time process simulation during execution
Predictive modeling for what-if scenarios
Virtual commissioning of new recipes
Continuous model refinement from execution data

Extended Reality Applications#

Augmented and virtual reality can enhance the paper on glass experience:

3D visualization of process equipment and flows
Immersive training environments for complex procedures
Remote collaboration on recipe development
Enhanced troubleshooting with contextual information overlay

Conclusion#

The paper on glass paradigm, when properly aligned with S88 and S95 ontologies, represents more than interface modernization—it creates a foundation for intelligent manufacturing. By maintaining operator familiarity while introducing semantic rigor, these systems bridge the gap between traditional process knowledge and modern digital capabilities.

The key to success lies not in abandoning proven workflows, but in enhancing them with structured data capture, standardized semantics, and intelligent automation. This approach preserves the valuable tacit knowledge embedded in traditional practices while enabling the data-driven insights necessary for competitive manufacturing.

As industries continue their digital transformation journeys, the paper on glass concept provides a pathway that respects operational heritage while embracing technological possibility. The alignment with established ontologies ensures that today’s digitization efforts build toward tomorrow’s intelligent manufacturing systems.