Upper ontology

Content sourced from Wikipedia, licensed under CC BY-SA 3.0.

Upper ontology (simplified overview)

What is an upper ontology?
- An upper ontology, also called a top-level or foundation ontology, is a very general framework for knowledge. It defines broad concepts like objects, properties, and relations that are common across many domains.
- Its main purpose is to help different domain ontologies work together. By providing a common starting point, it makes it easier to combine and compare data from different areas.

Why use an upper ontology?
- It supports semantic interoperability, allowing systems from different fields to understand and exchange information.
- Domain ontologies are built on top of the upper ontology, with their terms falling under the general categories defined there.

Development and challenges
- There is no single universally accepted upper ontology. Different groups have created their own versions, each with its own focus.
- Examples of well-known upper ontologies include Basic Formal Ontology (BFO) and Descriptive Ontology for Linguistic and Cognitive Engineering (DOLCE). Some are designed to be minimal and broadly applicable; others are larger and more detailed.
- Reasons why widespread agreement is hard include the complexity of the ideas, differing goals among communities, and political or organizational factors. Some critics argue that a single universal upper ontology may be infeasible, while supporters argue that a practical common ontology—built from a small set of primitive concepts—could be possible.

Arguments about feasibility

Arguments that upper ontologies may be infeasible
- There might never be universal agreement on a single set of concepts.
- Even if different representations exist, they might seem incompatible, making interoperability hard to achieve in practice.
- Some worry about definability: very general concepts can be hard to define precisely, and logic and definitions may be influenced by language, culture, and politics.

Arguments that upper ontologies may be feasible
- If we agree on a small core of primitive concepts, many domain ontologies can be built as combinations of those basics, enabling broad interoperability.
- An upper ontology can act as a translation layer, allowing different domain ontologies to communicate by mapping their terms to common primitives.
- Multiple views can be included as extensions or alternative theories within a single, coherent framework.
- Domain ontologies can be expanded by adding new primitives when needed, scaling up interop gradually.

How upper ontologies relate to real systems
- They are not meant to replace domain knowledge but to provide a common vocabulary and structure that domain ontologies can reference.
- Words and labels in ontologies are labels for concepts; the real power comes from the logical relationships and axioms, not the words themselves.
- Upper ontologies can help with data integration, standards, and machine-supported reasoning across domains.

Common upper ontologies (brief, easy-to-understand descriptions)
- Sowa’s Ontology: a compact, logically grounded foundation proposed by John F. Sowa.
- Cyc: a large, long-running project aiming to capture everyday common-sense knowledge; historically proprietary, with open versions available in the past.
- YAMATO: focuses on advanced descriptions of quality, attributes, properties, quantities, representations, processes, and roles; used in various applied ontologies.
- Basic Formal Ontology (BFO): a widely used top-level ontology designed to help domain ontologies interoperate; used in many biology and information-technology contexts.
- gist: a minimalist upper ontology intended for enterprise information systems; openly available and simple to apply.
- BORO: an extensional, four-dimensional ontology that emphasizes clear identity criteria and has influenced several standards.
- CIDOC Conceptual Reference Model (CRM) Core: a domain ontology for cultural heritage that includes a generic upper-layer subset for interoperability.
- COSMO: a collaborative effort to create a foundation ontology that supports broad semantic interoperability; designed to translate between different domain ontologies.
- ISO 15926: a standard for representing plant life-cycle information; a generic model with many entity types, used in industrial contexts.
- SUMO (Suggested Upper Merged Ontology): a comprehensive open upper ontology that links to many domain ontologies and to WordNet.
- UMBEL: Upper Mapping and Binding Exchange Layer, about 28,000 reference concepts; designed to connect OpenCyc, WordNet, DBpedia, GeoNames, and other sources.
- WordNet: a large lexical database used in language processing; it is not fully axiomatized as an upper ontology but provides broad semantic relations and is widely used in NLP.
- Other notable projects include DOLCE (Descriptive Ontology for Linguistic and Cognitive Engineering) and GFO (General Formal Ontology), each with their own emphasis and applications.

Key idea
- The main promise of an upper ontology is to provide a small, shared set of basic concepts that can be combined to describe many domain-specific ideas. This makes it possible to translate between different domain ontologies and to build interoperable knowledge systems.
- While debate continues about whether a single, perfect upper ontology will emerge, many researchers and organizations agree that using an upper ontology to standardize essential primitives can greatly improve cross-domain communication and data integration.

In short
- Upper ontologies aim to unify how we talk about the most general concepts so that diverse ontologies can work together.
- There are strong arguments both for and against a universal solution, but practical benefits come from agreeing on core primitives and building interoperable domain ontologies on top of them.