In the practice final exam solution for question #6 (see [[http://www.yioop.com/?c=group&a=groupFeeds&just_thread=559]]), it states: "An ontology is a representation of the objects in a domain, including facts about the world, but does not include the relations between the objects". In contrast, the textbook on page 308 defines an ontology as deciding on a "vocabulary of predicates, functions, and constants." What is more, the book continues, "The ontology determines what kinds of things exist, but does not determine their specific properties and interrelationships." A couple of us are confused on this point especially given that the example ontology in the practice final exam answer shows interrelationships between "things" as a type of hierarchy. Their solution is more in line with the definition of a general ontology on page 8 which says "a general ontology...organizes everything in the world into a hierarchy of categories." We were hoping for clarification on this point as we are unsure how to reconcile what appear to be somewhat contradictory definitions. What is more, in the same post for practice final exam question #6, it defines a taxonomy as: "A classification that can also have multiple sub-classifications." Our understanding is that a taxonomy is not specifically a classification but rather illustrate how subclass relations organizes into categories (see page 440). Any clarification here would also be much appreciated.

-- Ontologies and Taxonomies

Sorry, the example isn't really that clear. The example we gave was meant to be interpreted in a kind of OOP sense. That is, it represents object composition. Like: an ATM has a screen. The ATM represents a type of thing (i.e., a "class"), not a specific thing (i.e., not an "object" of the aforementioned "class"). We're not talking about relationships between things because we're not being that specific, we're talking about types of things and relations between types. That is, you know all ATMs have a screen, but you don't know which ATM has what screen. My understanding of it is that an ontology is an abstract specification of object types (of some form, I don't think it needs to involve the kinds of relations we used) you have in a domain as opposed to a concrete specification of the properties/relations of specific objects in that domain. By classification we mean category and by sub-classification we mean subclass relation. Probably should've used the book terminology there. But yeah, that's not well worded. It should be something like "a collection of subclass relations between categories". That's probably the most concise way to put it.

-- Ontologies and Taxonomies

The main place where this is talked about in the book is in Chapter 12, the definition on page 308 is sort of off-the-cuff. The word representation I think is confusing you. You can think of an ontology as specifying the kind of things that exist in a given "toy world" and how they are built up in terms of other events, time, physical objects, beliefs, etc in our toy world. So, for example, on page 442 of the book, they give an example of define a composite object Biped(a) in terms of other objects Leg(x) and Body(x). It does not give a particular biped "a" or give any relationship between biped "a" and another biped "b". Another thing you might specify in that one kind of thing is a subcategory of another, like Employee(x) is a subcategory of Person(x). When you give an ontology you are specifying a set of concepts like Biped, Person, Employee and on their definitions. An ontology that only defines thing using notions of subcategory, not composition, etc, is a taxonomy. Once we have specified the kinds of things can exist, then if we are given thing, we can start reasoning about it using our definitions and we can do this in a computational way since everything is completely spec'd out. Hope this helps.