Lecture
Ontology in computer science (New Latin ontologia, from Ancient Greek ὤν, genitive ὄντος — being, that which exists, and λόγος — study, science) is an attempt at a comprehensive and detailed formalization of some domain of knowledge by means of a conceptual scheme. Such a scheme usually consists of a data structure containing all the relevant classes of objects, their relationships, and the rules (theorems, constraints) adopted within that domain. In computer science this term is derived from the ancient philosophical concept of «ontology».
Ontologies are used in the programming process as a way of representing knowledge about the real world or part of it. The main areas of application are business process modeling, the Semantic Web, and artificial intelligence.
In information science, a top-level ontology (also called an upper ontology, TLO) is an ontology that includes very general terms — for example, «object», «property», «relation» — relevant to all subject domains. An important function of an upper ontology is to support semantic interoperability among a large number of domain-specific ontologies. This support involves creating a common starting point for formulating definitions. The terms of domain-specific ontologies are subordinate to the terms of the upper ontology.
Several projects for upper ontologies have been proposed. Any upper ontology can be regarded as a computational implementation of natural philosophy — the empirical method of inquiry within the ontology of the physical world .
Library classification systems anticipated the emergence of upper ontologies. Although library classifications organize and categorize knowledge on the basis of general concepts applicable to all subject domains, the systems do not replace one another.
Although the term «ontology» is originally philosophical, in computer science it has taken on an independent meaning. There are two essential differences here:
Modern ontologies are built in largely the same way, regardless of the language in which they are written. Usually they consist of instances, concepts, attributes, and relations.
Instances (or individuals) are objects — the basic, lowest-level components of an ontology; they can represent physical objects (people, houses, planets) as well as abstract ones (numbers, words). Strictly speaking, an ontology can do without specific objects, but since one of the main goals of an ontology is the classification of such objects, they are included as well.
Concepts or classes are abstract groups, collections, or sets of objects. They can include instances, other classes, or combinations of both. Example:
The classes of an ontology form a taxonomy — a hierarchy of concepts based on the nesting relation .
Objects in an ontology can have attributes. Each attribute has at least a name and a value and is used to store information that is specific to an object and bound to it. For example, the object Car-Model-A has attributes such as:
An attribute's value can be a complex data type. In this example, the value of the attribute called Engine is a list of values of simple data types.
An important role of attributes is to define relations (dependencies) between objects in an ontology. Usually a relation is an attribute whose value is another object.
Suppose an automobile ontology contains two objects — the cars Car-Model-A and Car-Model-B. Let Car-Model-B be a successor model of Car-Model-A; then we define the relation between Car-Model-A and Car-Model-B as the attribute «isSuccessorOf» with the value «Car-Model-A» for the object Car-Model-B (it should be noted that ontology description languages have predefined inheritance relations).
Specialized (domain-specific) ontologies represent some area of knowledge or part of the real world. Such an ontology contains term meanings that are specific to that domain. For example, the word «field» means a plot of land in agriculture, a type of matter in physics, and a class of algebraic systems in mathematics.
General ontologies are used to represent concepts common to a large number of domains. Such ontologies contain a basic set of terms, a glossary or thesaurus, used to describe the terms of subject domains.
If a system using specialized ontologies evolves, their merging may become necessary. A subtask of ontology merging is the task of ontology mapping. For an ontology engineer this is a serious challenge. Even ontologies of closely related domains can be incompatible with one another. The difference can arise from local cultural or ideological features, or from the use of a different description language. Ontology merging is performed both manually and semi-automatically. On the whole this is a labor-intensive, slow, and costly process. Using a base ontology — a single glossary — simplifies this work somewhat.
An ontology description language is a formal language used to encode an ontology. There are several such languages (the list is not exhaustive):
Several kinds of tools exist for working with ontology languages: ontology editors (for creating ontologies), ontology DBMSs (for storing and querying ontologies), and ontology repositories (for working with multiple ontologies).
Historically, many attempts in many societies have been made to impose or define a single set of concepts as more primary, fundamental, foundational, authoritative, true, or rational than all the others. A common objection to such attempts points out that people lack the transcendent viewpoint — or a God's-eye view — that would be required to achieve this goal. People are bound by language or culture, and therefore lack the objective standpoint from which the entire territory of concepts could be surveyed and any single standard derived. Thomasson , under the heading «1.5 Skepticism about category systems», wrote: «category systems, at least as they are traditionally presented, presuppose that there is a single correct answer to the question of what categories of entities exist — discovering that answer is the goal of most such research into ontological categories [...] But the actual proposed category systems differ so much that even a brief survey of past category systems like the one above can undermine the belief that there is such a unique, true, and complete system of categories. Given such a diversity of answers to the question of what ontological categories are, by what criteria could we choose among them to determine which one is uniquely correct?»
Another objection concerns the problem of formulating definitions. Upper-level ontologies are meant to provide maximal support for the functional interoperability of a large number of terms. Accordingly, such ontologies must consist of terms expressing very general concepts, but such concepts are so fundamental to our understanding that they cannot be defined, since the very process of definition implies that a less fundamental (and less understood) concept is defined in terms of concepts that are more basic and therefore (ideally) more understandable. Very general concepts can often be clarified, for example, through examples or paraphrase.
Those [ who? ] who doubt the feasibility of general-purpose ontologies are more inclined to ask: «What specific purpose do we have in mind for this conceptual map of entities, and what practical difference will this ontology make?» This pragmatic philosophical stance abandons all hope of developing an encoded ontological version of «The world is everything that is the case». (Wittgenstein, Tractatus Logico-Philosophicus).
Finally, there are objections similar to the objections against artificial intelligence . The technically complex acquisition of concepts and the social/linguistic interactions of people suggest that any axiomatic foundation for the «most basic» concepts must be cognitive-biological, or otherwise difficult to characterize, since we have no axioms for such systems. From an ethical standpoint, any universal ontology could quickly turn into an outright tyranny, recruiting adherents to a political program designed to spread it and to fund it, and possibly to defend it by force. Historically, inconsistent and irrational belief systems have proven capable of demanding obedience to the detriment or harm of people both within and outside the society that adopts them. How much more harmful would a consistent, rational one be, if it contained even one or two core tenets incompatible with human life?
Many of those who doubt the possibility of reaching broad agreement on a common upper ontology fall into one of two traps:
Indeed, different representations of statements about the real world (though not philosophical models), if they accurately reflect the world, must be logically consistent, even when they focus on different aspects of the same physical object or phenomenon. If any two statements about the real world are logically incompatible, one or both must be false, and this is a matter for empirical investigation, not for ontological representation. In practice, representations of the real world are created as, and known to be, approximations of the underlying reality, and their use is limited by the margin of measurement error in any given practical application. Ontologies are fully capable of representing approximations, and are also capable of representing situations in which different approximations have different degrees of usefulness. Objections based on people's differing perceptions of things attack an oversimplified, impoverished view of ontology. The objection that logically incompatible models of the world exist is true, but within an upper ontology these different models can be represented as different theories, and adherents of these theories can use them in place of other theories while preserving logical consistency in thenecessary assumptions of the upper ontology. The necessary assumptions provide the logical vocabulary , by means of which the meanings of all incompatible models can be defined. It has never been demonstrated that incompatible models cannot be properly defined using a common, more basic set of concepts, whereas there are examples of incompatible theories that can be logically defined only by means of a few basic concepts.
Many objections to upper ontology concern problems of life-critical decisions or non-axiomatized problem domains, such as law, medicine, or politics, which are difficult even for people to understand. Some of these objections do not apply to physical objects or standard abstractions that are created by people and carefully controlled by them for mutual benefit — for example, standards for connecting to electrical power systems, or the signals used at traffic lights. No single unified metaphysics is needed in order to agree that some such standards are desirable. For instance, although time and space can be represented in different ways, some of these representations are already used in interoperable artifacts such as maps or schedules.
Objections to the feasibility of a general upper ontology also fail to take into account the possibility of reaching agreement on an ontology containing all the primitiveelements of an ontology, which can be combined to produce any number of more specialized representations of concepts. Adopting this tactic makes it possible to focus effort on agreeing upon only a limited number of ontology elements. Having agreed on the meanings of this basic list of concepts, it becomes possible to create, and then precisely and automatically interpret, an infinite number of representations of concepts as combinations of the basic ontology elements. Any domain ontology or database that uses the elements of such an upper-level ontology to define the meanings of its terms will be automatically and precisely interoperable with other ontologies that use the upper ontology, even if each of them separately defines a large number of domain elements not defined in the other ontologies. In that case, correct interpretation will require that the logical descriptions of the domain-related elements be transmitted along with any data being exchanged; the data will then be interpreted automatically, because the descriptions of the domain elements, based on the upper ontology, will be properly interpreted by any system that can correctly use the upper ontology. In essence, elements in different domain ontologies can be *translated* into one another using the common upper ontology. An upper ontology based on such a set of primitive elements can include alternative representations, provided that they are logically compatible. Logically incompatible models can be represented as alternative theories, or presented as a specialized extension of the upper ontology. The proper use of alternative theories is itself a piece of knowledge that can be represented within the ontology. Users who develop new domain ontologies and find that there are semantic primitives needed for their domain but missing from the existing common upper ontology can add these new primitives through an accepted procedure, extending the common upper ontology as needed.
Most proponents of upper ontology argue that several good ontologies can be created, possibly with different emphases. Very few actually claim to be uncovering the one and only language, whether in natural language or even within an academic field. Most of them simply standardize some existing forms of communication. Another view that has been put forward is that there is almost complete overlap among the different ways of formalizing upper ontologies, in the sense that different ontologies focus on different aspects of the same objects, but different perspectives complement one another rather than contradicting each other; as a result, an internally consistent ontology is possible that contains all the representations along with means of translating the different representations into one another. However, such an ontology has not yet been built, because a large development project would be required to incorporate all the alternative views into separately developed upper ontologies together with their translations. The main obstacle to building such an ontology is not technical problems, but the unwillingness of funding organizations to provide resources for a sufficiently large consortium of developers and users.
Several common arguments against upper ontology can be more clearly examined by separating questions of definition (ontology), language (lexicons), and facts (knowledge). For example, people have different terms and phrases for the same concept. However, this does not necessarily mean that these people have different concepts in mind. They may simply be using a different language or idiom. Formal ontologies usually use linguistic labels to denote concepts, but the terms denoting elements of an ontology mean no more and no less than what they mean in their axioms. Labels are similar to variable names in software, but they are more suggestive than defining. Proponents of a common upper ontology point out that the meanings of elements (classes, relations, rules) in an ontology depend only on their logical form, not on the labels, which are usually chosen simply to make the ontology more convenient for human developers to use. In fact, labels for elements in an ontology do not necessarily need to be words — they could, for example, be images of instances of a particular type, or a video of an action represented by a particular type. It cannot be overstated that words — are *not* what is represented in an ontology, but rather entities in the real world, or abstract entities (concepts) in people's minds. Words are equivalent not to elements of the ontology, but to the words that *label* the elements of the ontology. There may be many words denoting a single concept even within one language (synonymy), and there may be many concepts denoted by a single word (ambiguity). Creating mappings between human language and the elements of an ontology is the domain of natural language understanding. But the ontology itself stands independently as a logical and computational structure. For this reason, reaching agreement on the structure of an ontology is actually easier than developing a controlled vocabulary, because all the various interpretations of a word can be included, each of which is «mapped» to the same word in different terminologies.
The second argument is that people believe different things and, therefore, cannot hold the same ontology. However, people can assign different truth values to a specific statement while accepting the validity of certain underlying statements, facts, or ways of expressing an argument they disagree with. (Using, for example, a question/position/argument form.) This objection to upper ontologies ignores the fact that a single ontology can represent different belief systems, as well as represent them as different belief systems, without taking a position on the validity or.
Even arguments about the existence of a thing require a certain separation of the concept, even if its existence in the real world may be disputed. Separating belief from naming and definition also helps clarify this issue and show how concepts can be shared even despite differing beliefs. For example, a wiki as a medium may allow such confusion, but disciplined users can apply dispute-resolution methods to resolve their conflicts. It is also argued that most people share a common set of «semantic primitives», fundamental concepts they refer to when trying to explain unfamiliar terms to other people. An ontology that includes representations of these semantic primitives could then be used to construct logical descriptions of any term a person might wish to define logically. This ontology would be one form of upper ontology, serving as a logical «interlingua» that can translate ideas in one terminology into their logical equivalent in another terminology.
Proponents argue that most disagreements about the viability of upper ontology can be traced to a confusion of ontology, language, and knowledge, or to overly specialized domains of knowledge: many people, or agents, or groups will have regions of their respective internal ontologies that do not overlap. If they can cooperate at all and share a conceptual map, this can be so beneficial that it outweighs any drawbacks arising from the sharing. To the extent that it becomes harder to share concepts the deeper one digs, the more valuable such an exchange becomes. If the problem is as simple as opponents of upper ontologies claim, then it also applies to a group of people trying to cooperate, who may need machine assistance to communicate easily.
At the very least, such ontologies are implied by machine translation , used when people are practically unable to communicate. Regardless of whether they are «upper» or not, they seem set to proliferate.
The Basic Formal Ontology (BFO) framework, developed by Barry Smith and his collaborators, consists of a series of sub-ontologies at different levels of granularity. Ontologies are divided into two kinds: those pertaining to continuant entities, such as three-dimensional enduring objects, and those pertaining to occurrent entities (primarily) processes, viewed as unfolding in successive phases over time. Thus BFO incorporates both three-dimensional and four-dimensional views of reality within a single framework. The relationships between the two types of ontologies are defined in such a way that BFO can deal with both the static/spatial and the dynamic/temporal characteristics of reality. A continuant domain ontology derived from BFO can be thought of as an inventory of entities existing at a given moment. Each occurrent ontology can be regarded as an inventory of processes unfolding over a given interval of time. Both BFO itself and each of its extension sub-ontologies can be viewed as a window onto a particular part of reality at a given level of granularity. More than 350 ontology frameworks based on BFO are catalogued on the BFO website. They apply the BFO architecture to various domains through a top-down population strategy. The Cell Ontology, for example, is populated downward from BFO by importing the branch of BFO ending in object and defining cell as a subtype of object. Other examples of ontologies extending BFO are the Ontology for Biomedical Investigations (OBI) and other ontologies of the Open Biomedical Ontologies Foundry . In addition to these examples, BFO and its extensions are increasingly used in the defense and security domains, for example in the Common Core Ontology framework. BFO also serves as the upper level of the Sustainable Development Goals (SDG) Interface Ontology, developed by the United Nations Environment Programme and the manufacturing industry's Industrial Ontologies Foundry (IOF) initiative. BFO was documented in the textbook Building Ontologies with Basic Formal Ontology , published by MIT Press in 2015.
The Business Objects Reference Ontology — is an upper ontology intended for developing ontological or semantic models for large, complex operational applications, comprising both an upper ontology and an ontology-building process. It is built on a number of clear metaphysical choices that provide a solid (metaphysical) foundation. A key choice was in favor of an extensional (and hence four-dimensional) ontology , which provides a simple criterion of identity . Its elements have appeared in a number of standards. For example, an early version of the ISO 15926 standard «Industrial automation systems and integration» was strongly influenced by it. IDEAS (International Defence Enterprise Architecture Specification for exchange) is based on the BORO standard, which in turn was used to develop DODAF 2.0.
Although the «CIDOC Object-Oriented Conceptual Reference Model» (CRM) is a domain ontology specialized for the purpose of representing cultural heritage, a subset called the CRM Core represents a general upper ontology, including:
A persistent item is a physical or conceptual object that has a persistent identity, recognizable throughout its existence through its identification, rather than through its continuity or observation. A persistent item is comparable to an endurant.
A propositional object is a set of statements about real or imagined things.
A symbolic object is a sign/symbol or a collection of signs or symbols.
COSMO (COmmon Semantic MOdel) is an ontology that was initiated as a project of the COSMO working group of the Ontology and Taxonomy Coordinating Working Group with the goal of developing a foundational ontology that could serve to provide broad, general semantic interoperability . The current version is an OWL ontology, but a Common-Logic-compatible version is expected in the future. The ontology and explanatory files are available on the COSMO website. The goal of the COSMO working group was to develop a foundational ontology through a collaborative process that would allow it to represent all the core ontology elements that all participants considered necessary for their applications. Development of COSMO is fully open, and comments or suggestions from any source are welcome. After some discussion and contributions from participants in 2006, development of COSMO was continued mainly by Patrick Cassidy, chair of the COSMO working group. Contributions and suggestions from any interested party are still welcomed and encouraged. Many types (OWL classes) in the current COSMO were taken from OpenCyc OWL version 0.78 and from SUMO. Other elements were taken from other ontologies (such as BFO and DOLCE) or developed specifically for COSMO. Initially, development of COSMO focused on including representations of all the words in the defining vocabulary controlled by the Longman Dictionary of Contemporary English (LDOCE) (2148 words). These words are sufficient to define (linguistically) every entry in LDOCE. It is assumed that ontological representations of the concepts represented by these terms will be sufficient to specify the meanings of any specialized ontology element, thereby serving as a foundation for general semantic interoperability. Interoperability through COSMO is achieved by using COSMO (or an ontology derived from it) as an interlingua, through which ontologies of other domains can be translated into each other's terms and thus communicate accurately. As new domains join COSMO, additional semantic primitives may be recognized and added to its structure. The current (January 2021) version of COSMO for OWL has more than 24,000 types (OWL classes), more than 1,350 relations, and more than 21,000 constraints. COSMO itself (COSMO.owl) and other related and explanatory files can be obtained via the COSMO link in the «External links» section below.
A well-known and fairly comprehensive ontology available today is Cyc , a proprietary system under development since 1986, consisting of a foundational ontology and several domain-specific ontologies (called microtheories ). A subset of this ontology has been released free of charge under the name OpenCyc , and a more or less complete version is available for free non-commercial use under the name ResearchCyc .
The Descriptive Ontology for Linguistic and Cognitive Engineering (DOLCE) — is a foundational ontology developed in 2002 in the context of the WonderWeb EU project by Nicola Guarino and his colleagues at the Laboratory for Applied Ontology (LOA). As its abbreviation suggests, DOLCE focuses on defining ontological categories underlying natural language and human common sense . DOLCE, however, does not adhere to a strictly referential metaphysics tied to the intrinsic nature of the world. Rather, the categories it introduces are treated as cognitive artifacts that ultimately depend on human perception, cultural imprinting, and social conventions. In this sense they are intended to be merely descriptive (rather than prescriptive ) notions that support the formal specification of a domain's conceptualization.
DOLCE-Ultralite , developed by Aldo Gangemi and his colleagues at the Semantic Technology Laboratory of the National Research Council (Italy), is an OWL (Web Ontology Language) version of DOLCE. It simplifies some of DOLCE's modal axioms and extends it to cover the descriptions-and-situations framework, also developed in the WonderWeb project. DOLCE-Ultralite is the source of some fundamental ontology design patterns [20] and is widely applied in ontology projects around the world.
The General Formal Ontology (GFO), developed by Heinrich Herre and his colleagues at the Onto-Med research group in Leipzig , is a realist ontology integrating processes and objects. It attempts to incorporate many aspects of recent philosophy, reflected both in its taxonomic tree and in its axiomatizations. GFO allows for different axiomatizations of its categories (for example, the existence of atomic time intervals versus dense time ). The core principles of GFO are published in Onto-Med Report No. 8 and in «General Formal Ontology (GFO): A Foundational Ontology for Conceptual Modelling». [21] [22]
Two features of GFO, among others, are its treatment of persistence and its temporal model. Regarding persistence, the distinction between endurants (objects) and perdurants (processes) is made explicit within GFO through the introduction of a special category, persistant. [23] A persistant is a special category whose instances are meant to «remain identical» (over time). Regarding time, time intervals are considered primitive in GFO, while time points (called «time boundaries») are derived. At the same time, points in time can coincide, which is convenient for modelling instantaneous changes.
gist is developed and maintained by Semantic Arts . gist (not an acronym — it means getting the gist) is a «minimalist upper ontology». gist is geared toward enterprise information systems, though it has also been applied to healthcare-delivery applications. The main attributes of gist are:
gist has been used to build enterprise ontologies for a number of major commercial and government agencies, including: Procter & Gamble, Sentara Healthcare, the Washington State Department of Labor and Industries, LexisNexis, Sallie Mae, and two major financial-services firms. gist is freely available under a Creative Commons license. gist has been actively maintained and used for more than 10 years. As of October 2020, it is at version 9.4.
gist has been the subject of a paper on how to overcome modelling differences between ontologies. In a paper describing the OQuaRE methodology for evaluating ontologies, the core unit-of-measure ontology (at the time a separate module) received the highest score in manual evaluation compared with 10 other units-of-measure ontologies and was rated above average in automated evaluation. The authors stated: «This ontology can be easily tested and verified, its knowledge can be efficiently reused and adapted for various specific environments».
Comments