Introduction to Ontologies
An ontology is a representation of the nature of things, and relationships things have between each other. The term ontology is derived from the greek words ontos (being) and logica (study of). Ontologies provide a formal mechanism to understand the nature of things. For instance we can make a lot of inferences about an apple by referring to existing ontologies. We can infer that it is a fruit of the Malus domestica in the rose family Rosaceae using an ontology called taxonomy of living things. We can infer that the apple consists of several complex molecules, which can be broken down into various elements that are represented in another ontology called the periodic table of elements. If we want to find the definition of an apple, we can look it up in a dictionary, a type of ontology known as a lexicon. All of these ontologies are commonly accepted representations of knowledge, and we can use them to make inferences about the things we encounter everyday in our lives.
2.3 Introduction to Ontologies
Download PDF
Introduction to ontologies
An ontology is a representation of the nature of things, and relationships things have between each other. The term ontology is derived from the greek words ontos (being) and logica (study of). Ontologies provide a formal mechanism to understand the nature of things. For instance we can make a lot of inferences about an apple by referring to existing ontologies. We can infer that it is a fruit of the Malus domestica in the rose family Rosaceae using an ontology called taxonomy of living things. We can infer that the apple consists of several complex molecules, which can be broken down into various elements that are represented in another ontology called the periodic table of elements. If we want to find the definition of an apple, we can look it up in a dictionary, a type of ontology known as a lexicon. All of these ontologies are commonly accepted representations of knowledge, and we can use them to make inferences about the things we encounter everyday in our lives.
Figure 1: Carl Linnaeus's taxonomy of living things
Similarly, we can create an ontology for a business domain which is a representation of things within the business. Within an enterprise there are usually several departments, such as the commercial, marketing, sales, operational, development, and support departments. Each of these departments can have their own perspective on the nature of things within the business. Each of these perspectives go towards building a complete ontology, but do not form an ontology of the business by themselves. A commonly accepted ontology within an enterprise would serve as a basis for making inferences about the business, just as we did with the apple. A business ontology can support the building of information systems and making business decisions.
Figure 2: Dmitri Mendeleev's periodic table of elements
Creating stable ontologies
Businesses are prone to change, and the construction of a business ontology must take into account the nature of the changes the business experiences. If we want to make a business ontology that withstands changes, we can learn a lot of lessons from 2 ontologies that have changed very little since their original design. The taxonomy of living things was originally developed by the Swedish naturalist Carl Linnaeus in the 18th century, and it is still used today based on its original principles. The taxonomy of living things primarily makes classifications on the basis the physical characteristics of living things. Other criteria, such as sexual-orientation, social structures, or diet could be used as the primary basis for classification, but it would yield very different results. As an example, consider the 2 different species of chimpanzee. The genus chimpanzee (Pan) has 2 species, the pygmy chimpanzee (Pan paniscus) and the common chimpanzee (Pan troglodytes). Neither species can swim, and it is speculated that the formation of the Congo river around 2 million years ago led to the separation of the same species on either side of the river. Pygmy chimpanzees have a matriarchal culture, are omnivorous, and engage in sexual behaviour with multiple partners. The common chimpanzees, on the other hand, have a patriarchal culture, are moslty herbivorous, and are usually monogamous. If the primary basis of classification were to be social behaviour, then the pygmy chimpanzee and the common chimpanzee wouldn't share the same genus. Social behaviour is usually prone to change, and using it as a basis to construct an ontology would mean that the ontology would have to be updated more frequently. Using the physical characteristics of a living being as the ontological key, or the primary basis on which an ontology is based, would create a far more stable representation.
Figure 3: The pygmy chimpanzee (left) and the common chimpanzee
Similarly, let's consider an ontology of an organization. Organizational charts are a common representation of an organization. The ontological key, or the basis upon which someone gets placed higher or lower on an organization chart, is function of power and responsibility. As both power and responsibility are prone to constant change within an organization, organizational charts are one of the most unstable ontologies. They are prone to constant change, and in some organizations, an organizational chart becomes outdated moments after it is updated. If we want to create stable ontologies, it is important to base them on stable ontological keys.
Figure 4: An organizational chart
A simple ontology
Let's consider an ontology of 2 birds, the Robin and the Bluebird. A Robin lays eggs, has feathers, breathes, and has skin underneath the feathers. A Bluebird also lays eggs, has feathers, breathes, and has skin underneath the feathers. Some of the characteristics of Robins and Bluebirds are common to all birds, namely that they lay eggs, have feathers, breathe, and have skin underneath the feathers. Some of the characteristics of birds are common to all animals, namely that they have skin and they breathe. In creating an ontology for the Robin and the Bluebird, we can create 2 abstractions, one for birds, and another for animals. We can place the characteristics of laying eggs and having feathers as the properties of all birds. Similarly, we can place the characteristics of having skin and breathing as the properties of all animals. There are a couple of benefits of creating these 2 abstractions. First, we can focus on the specific properties of Robins and Bluebirds, reducing the amount of information we have to manage for each bird. Next, the abstractions allow us to view the characteristics of all birds and deal with them in a common way. For instance, if we wanted to innoculate all birds against the bird flu, we could simply deal with the information model for birds, and not have to deal with robins, ducks, bluebirds, teals, penguins, etc.
Figure 5: A simple ontology based on the Robin and the Bluebird
Beware of making abstractions on limited information
A downside of creating abstractions is that they are often invalid if they are not built bottom-up. Here, we have built a simplistic model based only upon 2 birds. This model may also lead to the conclusion that any animal that lays eggs is a bird, but reptiles also lay eggs. We could have placed the property of flying as a property of all birds based on our sample set of the Robin and the Bluebird, but not all birds can fly (e.g. the Penguin, Emu, and the Dodo). Abstractions should be created very carefully and by studying a large sample set. As a final cautionary note, consider the platypus, which completely baffled naturalists when it was first found, as it couldn't be classifed as a bird (it had a beak) or a mammal (it laid eggs). It was also venomous, a characteristic of reptiles, had a tail like a beaver, and feet like an otter. When it was initially discovered, the majority of naturalists thought it was a hoax. Eventually, a new category of mammals was created called monotremes where Platypuses could safely be placed.
Figure 6: The Platypus
Download PDF
