14 July 2005

Ontological Components: Entities

The ontology of an information system (whether for computer science, cognitive science, information management or other areas) differs somewhat from the Ontology referred to by philosophy. In philosophy, Ontology is the theory of being [reference], it is the theoretical description about all that is, and how it relates to each other, within a universe of discourse. There is a subtle difference existing within the study of ontology for purposes of enabling an information system - a difference that is helpful to keep in mind when considering entities. That difference is this - in an information system, ontology is the study of the REPRESENTATION of all that exists within the universe of discourse.

Taxonomies, Knowledge Bases, and Ontology

An interesting side topic that is of value to consider when discussing entities and the method for evaluating them is this: an ontology is not a taxonomy, nor is it a knowledge base, but it can include both of those things.

A taxonomical structure is a hierarchical classification of all the data or linguistic objects (words, datum, elements, etc.) to be found within a universe of discourse [reference]. A knowledge base is a taxonomy (which is just a hierarchy of classes) that has been populated with enumerations matching the various classes [reference]. A formal ontology includes the former, among other things, but where the elements of a taxonomy differ when they are included in a formal ontology is this - an ontological view of a taxonomy also includes a definition of all the underlying assumptions and properties that define each of the classes in the taxonomy. A formal ontology need not contain the enumerations that exist within a knowledge base (and it seems difficult to imagine such an enumeration being complete in any but the most simple of universes of discourse). As I mentioned this is a side topic, but it is interesting to know what we conceive of when we mention the terms taxonomy and knowledge base.

The entities of a formal ontology include the classified and defined representations of objects, events, and phenomena. Each of these are classified in a hierarchy similar to a taxonomy. They are defined by their component concepts, for all entities are compounds (as defined in the previous section). Some of the characteristic properties of entities derive from their place on the taxonomical hierarchy, and these should be captured as concepts of the entity.

Entity Class and Entity Instance

We see therefore that entities within a formal ontology are classes, in the taxonomical sense, but with a formal definition of how those classes are defined (the exhibition of concepts related to the entity). This (the class) is one of two possible views of entities within a formal ontology. The second possible view of an entity is the potential for enumerations within that class.

It is possible for an enumerated entity to begin to have a separate identity from that of its parent class. This is through the introduction of non-persistent properties. Each enumerated entity shares the same persistent properties, and these are defined by the taxonomical class, as well as all the ontological concepts that define that class. All enumerations deriving from that class inherit those concepts. These are the persistent properties of those enumerations. However, the nature of an entity to exist within space and time, and an entity's nature to change state as it is related to other entities (particularly phenomena), determine that it will have some properties that are non-persistent, and these change not only from enumeration to enumeration, but also over time.

As an example, let us consider the entity "GP40 Diesel Electric Locomotive". This entity is a class within a formal ontology that considers (as part of it's universe of discourse) railroad assets. Some of the persistent properties of this class are these - 3000hp engine, 40 inch wheel diameter, rated weight 257,000 lbs, etc. These are the properties common to the whole class. Now, if we take a particular enumeration of this class, say "CSX engine number 6936", this enumeration takes on a number of non-persistent properties. It's current location (Louisville KY?), it's consist of freight cars, it's driver, it's current paint job, it's maintenance state, it's actual rate. All of these things are properties that can define such an entity, if the ontological need exists. The value of these properties change (or perhaps not be existent) when considering other enumerations of the entity "GP40 Diesel Electric Locomotive". Property exhibiting concepts might exist across the various enumerations, but the value of those properties might change (such as location - each existing instance of a GP40 has the concept location, but the value of that concept changes from instance to instance, and over time).

It is important to make the distinction between the enumerations of a particular taxonomical class, and separate, but related classes. This distinction becomes especially important as we consider hypernyms and hyponyms within a taxonomical hierarchy. Hypernyms and hyponyms are redefinitions of classes, into other classes, at different resolutions. This is easier to see through example, than it is to explain. For instance, the taxonomical class "vehicle" is a hypernym for the classes "chariot", "helicopter", "locomotive" and others. Those other classes are also vehicles, but the more precise class is considered at a higher resolution of detail. Likewise, the taxonomical classes "hydrocodone", "penicillin", and "ritalin" are all hyponyms for the taxonomical class "drug". The class "drug" is similar to the others, but considered at a lower resolution of detail.

In a graphical representation showing the hierarchy of the classes of a taxonomy, the hypernyms are parents and the hyponyms are children. The failings of this sort of representation are that it becomes very complicated very quickly, when you have classes that are hyponyms for multiple "parents", which is very easy within any but the most simple of taxonomies (or ontologies).

The Categories of Entity

Within a formal ontology, there is room for both aspects of an entity, the class and the instance (or, enumeration). The class comes from the roots that a formal ontology has in a taxonomy, and the presence of instances allow the formal ontology to serve as a knowledge base. Both of these are required, as both have separate property issues that need to be addressed by the ontological concept structure. The same is true for all three aspects of entity - object, event, and phenomena.

  1. Objects are easy to comprehend, as they are the objects and things of the world (universe of discourse) for which the formal ontology is describing representations. What is important to understand about objects is that they have several aspects that might not be immediately apparent.

    1. First, objects can tangible or abstract. By naming an object tangible, I mean that it can be a physical object (such as a truck, or a country). Equally valid, however, is that an object can be an abstract thing - something that has definite properties but is not physical. Some examples of abstract objects could include culture, organizations, or decisions.

    2. A second aspect to consider is that when discussing either tangible or abstract objects, those objects do not necessarily need to be "real". It is perfectly valid to have a formal ontology define representation of an entity (object or otherwise) that is non-real, so long as the entity has an "essence" that can be defined through a combination of property exhibiting concepts. This idea of non-real objects can include future objects, nominal objects, or potential objects (and in all cases, they can be tangible or abstract). The specific description of the representation is what is important for a formal ontology to be sound, not necessarily that all of the objects represented be soundly based in reality.

  2. In addition to objects, there is also the sub-category of entity, which we refer to as events. We define an event as "an entity with a time component". An event-entity is similar to an object-entity in that it can be tangible or abstract, but it has a period of time (which, as above with object-entities, need not be a real period of time, but could be future, past, nominal, or potential) during which it represents an entity with defined properties. An example of a tangible event is a rainstorm (with a start time, a stop time, and a tangible object related to it - rainfall). An example of an abstract event is a meeting (an ad hoc organization structure with a time component).

  3. The final sub-category of entity is phenomena. Phenomena are the entities that, when related to other entities through the relationship component of a formal ontology, change the state of the entity they are related to. This is done by affecting the properties or property-values of the affected entity. Phenomena share the aspects of entities in that they can be real or non-real, they can be tangible or abstract, and they can exist over time, similar to an event. Phenomena include the elements of linguistics that we think of as verbs and modifiers. Anything that implies action or change is a phenomena-entity, and some examples are damage, movement, unloading, growth, decay and others. They are related through relationships, and have properties (as all entities do) so that rules can be formed about their applicability for these relationships.

A Method for Entity Evaluation

Our method for evaluating entities must, as with concepts, be based in the intended use of the formal ontology. From that starting basis, we can move to the various aspects required for consideration in our evaluation.

  1. First entities must have properties (more accurately, property exhibiting concepts). The concepts defining entities, and giving them accessible properties, must be apparent and accessible. In fact, for ontological purposes, they must be defined. This is true not only of entity classes, but also of entity instances. It is equally true for not only properties but also property values. In the previous section on concepts, we defined the range of concepts of a compound to be all of the concepts that define that compound - in the case of an entity (which has two possible states of existence - as a class and as an instance), this range can exist in several different states. All of them must be explicitly addressable and apparent for the entity component to be evaluated to be adequate.

  2. The second consideration in our evaluation is the consideration of all the possible entities (objects, events, and phenomena). Are all of the requirements of communications within the universe of discourse satisfied by the enumerated list of all possible entity classes? Are the definitions of the entity instances sufficient to accommodate the needs of the universe of discourse?

  3. The final consideration is the depth of definition that the ranges of concepts provide in defining the entities. Are the entities defined to enough (and not too much) detail to afford the sorts of use they will be put to in the universe of discourse? If we have a universe of discourse that is discussing the movement of cargo through a supply system, then it is necessary for the entity "truck" to have the concepts of capacity, reliability, speed, ease of use, etc as property exhibiting concepts within the formal ontology. The concept of "what color is the seat inside the truck" is probably too much detail. But the lack of "how easy is the cargo bed to access" might be necessary.

Entities within a Fractal Ontology

A word about this final consideration is in order. There exists the idea of a fractal ontology (or "fractology" as a colleague has suggested recently), which implies that the level of examination that the entities and relationships within the formal ontology might change in resolution, depending on the use that it is being put to. To support this sort of idea, then the attendant concepts of entities must exist to appropriately support the highest and lowest resolution of consideration, and all levels in between that may be adopted. At that point, we have a formal ontology that has a dynamic resolution, but if the properties and concepts existing at those different levels are compounds or components of each other as the scale of consideration shifts, then the formal ontology becomes a fractal ontology.

Earlier article on Entities



Post a Comment

<< Home