Knowledge Engineering with Ontological Prisms (KEOPS)

*The shaping of Knowledge (Parallactic instrument of Taqî al-Dîn, Istanbul Topkapi Library*

[The book is available on Amazon and the O’Reilly digital platform.]

Outline

Iterative & Incremental Engineering

Taking for granted that knowledge is inherently dynamic, pushing edges across contexts and concerns, its engineering cannot be circumscribed in silos and time capsules. To meet the challenge, Knowledge Engineering with Ontological Prisms (KEOPS) combines iterative and incremental development.

KEOPS combines iterative (left) and incremental (right) moves.

Iterative moves traverse extensional, intensional, and logical realms. Extensional iterations consider native or man-made objects or phenomena that can be observed in symbolic or physical environments and are deemed relevant. Intensional iterations pertain to concepts that cannot be substantiated and consequently rely on ontological commitments addressing the mapping of concepts to facts. Logical iterations address the representations of selected extensional or intensional elements, using formal, and empirical languages.

Incremental moves operate like Jacquard looms, which are machines used to weave complex patterns and motifs into the fabric of textiles. Three rods, one for each realm, are moved parallel to the boundaries of the prisms. Inward moves introduce new facts, categories, or concepts from nondescript entries and use intersecting gears to map these new entries to existing ones. Outward moves translate successful mappings of facts, categories, and concepts into interoperable data, information, and knowledge, respectively.

Interoperability

The game-changing benefit of this double-edged approach is to address the integration of heterogeneous representations at different stages of ontologies life-cycle, obtained through the interoperability of realms: perceptions (alignment of meanings between observations and intentions), projections (interoperability of temporal and organizational representations), and experience (feedback and learning from doing).

Engineering Interfaces

The integration of heterogeneous and diachronic representations an be realized through established technical and functional interfaces.

Engineering Phases

Cycles of Iterative and incremental work units are structured by realms and synchronized through overlaps.

Mapping Territories: Facts & Modalities

While knowledge is not necessarily driven by purposes, it nevertheless aims at territories and topics. The first step is therefore to identify individuals deemed to be relevant within the territory considered. Such individuals can be directly classified according to their nature, whether physical or symbolic and characterized in terms of intrinsic identity (#) and functional aspects (≈). They can be further characterized by ontological modalities, e.g., instanciation, identification, or agency. Perceptions and experiences are then introduced to integrate individuals with existing concepts and categories, respectively.

Mapping Intensions: Concepts & Commitments

Contrary to models, which are meant to represent elements within specific and bounded contexts and concerns, ontologies are agnostic about reality. This absence of actual boundaries must be compensated by discretionary truth models and explicit commitments. Such models are made of truth-bearers (nodes) and truth-makers (connectors), which are subject to three kinds of commitments: semantic, with regard to environment perceptions, conceptual, with regard to the consistency of intensions, and logical, with regard to the consistency of projections.

Mapping Representations: Categories & Developments

While systems engineering aims to deliver packaged products in stable versions, knowledge engineering deliverables are rarely structured and are inherently subject to change and expansion in both content and scope. When used to support systems engineering, this issue can be managed by leveraging the distinction between information, data, and knowledge within ontological prisms, allowing them to be managed independently. However, this is not an option for the engineering of heterogeneous knowledge, which must continuously ensure the consistency of facts, concepts, and categories, as well as the operational integration of experiences, perceptions, and projections. Taking into account the amorphous nature of knowledge engineering outcomes, KEOPS adopts a reverse perspective and defines the results of knowledge engineering through use cases.

These use cases are first defined by scope and homogeneity:

Bounded domains ensure built-in consistency of words (facts), and meanings (concepts), while unbounded ones do not.
Homogeneous realms ensure built-in consistency of abstraction semantics, while heterogeneous ones do not.

Knowledge engineering use cases are further defined in terms of commitments regarding the kind of changes being considered. This shift from requirements (an imperative approach) to commitments (a declarative approach) is at the core of issues related to transparency, traceability, and accountability.