Knowledge Representation Questions and Answers

Answer: A

Writing the concept of things having some child who is a doctor uses an existential restriction on the has child role, while the universal restriction says every child is a doctor, which is a different and weaker claim satisfied vacuously by the childless. Concept non-emptiness is a separate assertion. Distinguishing some from every, and noticing that the universal form is satisfied when the role is empty, is the commonest source of modelling error in these languages.

Answer: B

A claim that every child is a doctor holds trivially of someone with no children, exactly as a universally quantified statement over an empty set is true. Modellers intending to require a child must add an existential restriction as well. Assumed roles, closed world reading and default exclusion are misreadings. A logical convention that is correct and counter-intuitive will be a persistent source of error unless it is stated explicitly during training.

Answer: C

Saying that a person has at most two biological parents, or a committee at least three members, uses a cardinality restriction on the relevant role. Such restrictions can render an ontology inconsistent when combined with assertions, which is precisely what makes them useful for detecting errors. Concept counts, domain size and inference limits are different notions. Constraints that can be violated are the ones that catch mistakes, which is why they are worth stating even when they seem obvious.

Answer: D

Inconsistency shows as a concept being subsumed by the empty concept, meaning nothing could possibly satisfy its definition, which usually indicates a modelling error rather than a discovery about the world. Direct contradictions are the simplest case of the same thing. Axiom counts and timeouts are unrelated. A check that turns a modelling mistake into a mechanically detectable condition is what makes large ontologies maintainable at all.

Answer: A

In an ontology of thousands of concepts a contradiction introduced by one axiom would never be found by reading, and a reasoner locates it in seconds. It says nothing about whether the ontology matches reality, which remains a matter of human judgement. Query correctness and review are not addressed. A mechanical check that catches a class of error completely, while saying nothing about correctness in the wider sense, is exactly as valuable as it is limited.

Answer: B

Categories are reified so that properties can be attributed to them, letting one say that the category of dogs is a subcategory of mammals or that it has many members. Membership and subclass relations then organise the whole knowledge base. Partitions, syntactic groupings and labels are different notions. Making a class into an object, so that facts about the class as such can be stated, is reification applied at the level of the taxonomy itself.

Answer: C

Fido is a member of the category of dogs, and dogs are a subclass of mammals, and confusing the two produces the classic error of treating a class as an individual. Subclass is transitive and membership is not: Fido is not a member of mammals by transitivity but by the combination of the two relations. Assertion, physicality and transitivity are misstated in the other options. Two relations that look similar and behave differently deserve careful separation in any formalism.

Answer: D

Declaring that animals are exactly the union of vertebrates and invertebrates, and that nothing is both, licenses inferences that neither subclass statement alone permits, such as concluding invertebrate from not vertebrate. Storage, file organisation and category counts are unrelated. Stating disjointness and exhaustiveness explicitly, rather than leaving them implicit in the modeller's intention, is what converts a taxonomy from documentation into something a reasoner can use.

Answer: A

Without a statement that cats and dogs are disjoint, nothing prevents an individual from being both, and no negative conclusion about class membership can be drawn. Modellers consistently underestimate how many such axioms are needed. Inconsistency, classification failure and indistinguishability do not follow. That the absence of an axiom weakens conclusions silently, rather than producing an error, is what makes omissions of this kind so hard to notice.

Answer: B

A defined concept states conditions that anything satisfying them must fall under, so individuals are classified automatically, whereas a primitive concept states only necessary conditions and membership must be asserted. The distinction determines whether the reasoner can infer membership or merely check it. Naming, enumeration and usage are different matters. Choosing between a definition and a primitive is the modelling decision that determines how much work the reasoner can do.