Subject-Event Ontologies

• Subject-event ontologies are formal frameworks that model relationships between subjects and events with rigorously defined roles, temporal intervals, and causal links.
• They enable diverse applications across NLP, process mining, distributed systems, and ethical reasoning by enforcing logical and domain-specific constraints.
• These ontologies support semantic interoperability and compositional event modeling, facilitating automated extraction, reasoning, and traceable knowledge aggregation.

Subject-event ontologies provide a formal framework for modeling the relationships between entities (subjects, objects, agents) and events within computational, linguistic, philosophical, and applied contexts. They underpin diverse applications in information extraction, process mining, distributed systems, and ethical reasoning by enabling rigorous specification of event types, participant roles, spatial-temporal constraints, causality, and compositionality.

1. Formal Foundations and Core Concepts

Subject-event ontologies universally distinguish between entities persisting through time (endurants) and occurrences extended over time (perdurants or events), with foundational categories formalized in upper ontologies such as DOLCE, UFO/gUFO, and variants. In DOLCE-based frameworks, events are “Perdurants” directly linked to time, while endurants/objects (DUL:Object, gufo:Endurant) act as event participants (Scherp et al., 2024, Hooshyar et al., 16 Dec 2025).

In one-category ontologies such as the Thinging Machine (TM or “thimac”), all modeled entities are “thimacs,” integrating the thing (subject) and machine (process) aspects. Here, an event is formalized as a pair $e = (R, \tau)$, where $R$ is a subgraph of the static machine representing potential action, and $\tau$ is the interval in which these potentials actualize, thereby encoding both the participating subject(s) and the temporal extent in a homogeneous formalism (Al-Fedaghi, 2022, Al-Fedaghi, 2020).

The relationship between subjects and events is encoded as a participation relation, formalized as $participatesIn(x, E)$ if $x$ is a node in the subgraph $R_E$ brought alive by time. Roles such as instigator (agent) or passive participant are explicit in most modern ontologies, with role-specific properties (e.g., hasAgent, hasPatient, hasParticipant) extending this basic participation (Borgo et al., 2016, Scherp et al., 2024, Shekarpour et al., 2017, Hooshyar et al., 16 Dec 2025).

2. Modeling Event Structure, Roles, and Time

Subject-event ontologies specify event types (classes), argument roles, temporal intervals, spatial regions, and compositional structure. The formalization uses Description Logics (DL), first-order logic, and/or RDF/OWL schema:

• Event Classes: E.g., Event, Action, Accomplishment, Process, State, with domain-specific specializations (e.g., Death, Homicide, CyberAttack) (Balali et al., 2021, Sobhani et al., 2019).
• Roles: Agent, Patient, Instrument, Tool, Source, Target, with ontology-specific refinements (e.g., MoralIntention, Consequence, NumberOfParticipants) (Aijaz et al., 7 Feb 2025, Shekarpour et al., 2017, Balali et al., 2021).
• Properties: hasAgent, hasPatient, hasParticipant, hasConsequence, upholdsEthicalPrinciple, hasTime, hasPlace.
• Temporal and Spatial Modeling: Events are anchored with TimeInterval, TimePoint ($t_s, t_e$) entities and associated via Allen-style relations (before, meets, overlaps). Spatial participation is encoded through geo:SpatialThing or DUL:SpaceRegion parameters (Scherp et al., 2024, Hooshyar et al., 16 Dec 2025, Vossen et al., 2017).
• Event Composition and Mereology: Ontologies like Event-Model-F provide explicit patterns for composition (Composite, Component), causality (Cause, Effect, Justification), correlation, and event identity at varying levels of granularity (Scherp et al., 2024, Vossen et al., 2017).

Multiple frameworks allow for explicit mereological (whole-part) and causal chains between events, supporting event aggregation, decomposition, and traceable provenance (Scherp et al., 2024, Hooshyar et al., 16 Dec 2025).

3. Subject-Event Participation and Role Assignment

The assignment of subjects to event roles is foundational for semantic interoperability and reasoning. In ontologies extending DOLCE or UFO/gUFO, subjects are reified as DUL:Objects or gufo:Endurants, linked to DUL:Event via participation properties. Advanced frameworks provide pattern-based role reification, subclassing generic Participant to domain-specific roles (e.g., FireFighter, Citizen, AffectedBuilding) and enabling multi-entity, multi-role linking (Scherp et al., 2024, Hooshyar et al., 16 Dec 2025).

Formally, ontologies assert for each event at least one participant: $$\forall e\;\bigl|\{o\mid \mathsf{hasParticipant}(e,o)\}\bigr|\ge1$$ and enforce cardinality and type constraints on role assignments. Ontology-specific role assignment axioms ensure that, for an event of subtype $s$, only roles in $AllowedRoles(s)$ are permitted (Balali et al., 2021).

Dynamic or context-dependent roles can be specified per event subtype, as in COfEE, enabling fine-grained representation (e.g., hasNumberOfDeaths, hasVehicle, hasInstrument) beyond generic agent/patient axes (Balali et al., 2021).

4. Event Composition, Causality, and Relation to Time

Subject-event ontologies extend modeling from simple event-participation to complex event composition, causal relationships, and temporal constraints:

• Mereology: Patterns for event composition (Composite/Component) allow for the specification of events as aggregates or subevents, supporting nested and hierarchical event structures (Scherp et al., 2024, Hooshyar et al., 16 Dec 2025).
• Causality and Correlation: Explicit classes and relations (Cause, Effect, Justification, Correlate) allow specifying direct causal chains and correlative associations among events. Acyclicity is usually enforced: causality is modeled as a transitive, irreflexive relation (Boldachev, 20 Oct 2025, Scherp et al., 2024).
• Temporal Reasoning: Time parameters are associated using DL axioms (e.g., TimeParameter ⊑ ∃parametrizes.DUL:TimeInterval), with Allen interval relations for event ordering (Hooshyar et al., 16 Dec 2025).

In subject-event ontologies without global time (e.g., Boldsea), causal order is defined purely by explicit dependencies (happens-before), not by timestamps, supporting distributed and asynchronous systems. Event validation, append-only histories, and multiperspective modeling are enforced through a set of axioms and execution semantics, guaranteeing monotonicity, acyclicity, and traceability (Boldachev, 20 Oct 2025).

5. Applications Across Domains

The application of subject-event ontologies demonstrates domain-agnostic flexibility:

• Information Extraction & NLP: Annotating and extracting event-centric knowledge from text via frameworks such as CEVO and COfEE, which provide cognitive event taxonomies and explicit subject–event–role triples mapped from natural language (Shekarpour et al., 2017, Balali et al., 2021).
• Process Mining and Object-Centric Analytics: gOCED, grounded in gUFO, enables explicit specification of objects, events, time, and reified dynamic relations (e.g., contracts, supervision), resolving ambiguities in classical process/event-centric models (Hooshyar et al., 16 Dec 2025).
• Distributed Systems: Execution semantics in ontologies like Boldsea allow for microservices and DLT platforms to coordinate by event causality, not global time, ensuring traceable, concurrent, multiperspective operations (Boldachev, 20 Oct 2025).
• Ethical Reasoning: The ApplE ontology bridges normative theories (consequentialism, deontology) and concrete event/action modeling, supporting automated ethical assessment in applied and biomedical scenarios (Aijaz et al., 7 Feb 2025).
• Surveillance and Forensics: The Forensic Event Ontology classifies complex criminal events, linking participants to actions for video analysis, event inference, and automated retrieval at various semantic levels (Sobhani et al., 2019).

6. Granularity, Event Identity, and Coreference

Event identity and granularity are interdependent: ontologies expose parameters (action alignment, temporal granularity, shared participants) that tune the strictness of coreference and clustering. The Simple Event Model (SEM) encodes subject-event relations and supports coreference, subevent, and topical relations, with semantic anomaly checks to prevent contradictory merges. Formally, identity is a thresholded sum of similarities across actions, times, participants, and places (Vossen et al., 2017, Borgo et al., 2016).

Rule-based reasoning frameworks utilize these models for coreference annotation, pre/post-condition inference, and event disambiguation, leveraging ontological constraints for precision and traceable knowledge aggregation (Borgo et al., 2016, Vossen et al., 2017).

7. Methodologies, Modularity, and Extension

Ontology development methodologies include pattern-oriented approaches (Event-Model-F), methodical role and type specification in DL or first-order logic, and cyclic agile methods for iterative formalization (as in ApplE's SAMOD-based design). Modularity is enforced by separating participation, composition, causality, correlation, and interpretation into distinct OWL/DL modules, promoting extensibility and reusability. Domain extension involves subclassing pattern roles and event types, binding new properties, and specializing constraints for application-specific semantics (Scherp et al., 2024, Aijaz et al., 7 Feb 2025).

Tooling and gold-standard datasets support language-independent annotation and benchmarking, as demonstrated in COfEE’s annotated Persian news corpus and its supervised extraction benchmarks (Balali et al., 2021). Pattern- and modularity-based frameworks facilitate domain adaptation and practical deployment.

---

The formal apparatus of subject-event ontologies, as implemented across foundational ontologies and domain-specific frameworks, provides a structured basis for semantic integration, reasoning, and automated processing of events, participant dynamics, time, causality, and provenance in diverse computational and knowledge-intensive domains (Scherp et al., 2024, Hooshyar et al., 16 Dec 2025, Aijaz et al., 7 Feb 2025, Boldachev, 20 Oct 2025, Al-Fedaghi, 2022, Balali et al., 2021, Shekarpour et al., 2017, Vossen et al., 2017, Sobhani et al., 2019, Al-Fedaghi, 2020, Borgo et al., 2016).