Knowledge Representation Questions and Answers

Answer: C

A TMS stores why each conclusion was drawn, so when a premise is withdrawn the conclusions depending on it can be retracted efficiently rather than by rebuilding everything from scratch. It is the practical machinery that makes non-monotonic reasoning affordable. Verification, consistency enforcement and timestamps are related but different. Recording the derivation as well as the result is what allows a system to change its mind without starting again.

Answer: D

Description logics restrict expressiveness deliberately so that subsumption, whether one concept necessarily includes another, remains decidable and often tractable. They underpin the web ontology language. Program logics, single object restriction and natural language are different notions. Choosing what to give up in order to keep reasoning decidable, rather than seeking maximum expressiveness, is the design philosophy that distinguishes this family.

Answer: A

Subsumption asks whether one concept description necessarily includes another, which arranges the concepts into a hierarchy computed rather than declared. Instance checking is the related question about individuals, and knowledge base merging and simplification are different operations. That the taxonomy is derived from the definitions rather than asserted by hand is the principal advantage of the approach and it keeps the hierarchy consistent automatically.

Answer: B

If the hierarchy is derived from the definitions, it is impossible for a concept to be placed under a parent it does not actually specialise, which in a large hand built ontology happens constantly. Multiple inheritance is supported by both approaches. Concept counts and speed are secondary. Deriving a structure from its definitions rather than maintaining it separately removes a whole class of error, which is the same argument that favours generated over hand written code.

Answer: C

An engineering ontology fixes the categories, the relations between them and the constraints they obey, so that several systems can exchange knowledge with a shared understanding. It borrows its name from the philosophical enquiry but is a practical artefact. Instance data populate it, and a dictionary lacks the formal structure. That interchange between systems requires agreement about categories, not merely about syntax, is what motivates the whole enterprise.

Answer: D

An upper ontology fixes the most general distinctions, such as between objects and processes or between abstract and physical things, so that domain ontologies can be built beneath it consistently. Agreement at that level has proved contentious. Single top concepts, software layers and version currency are different notions. Whether general agreement about the highest categories is achievable at all is a genuine open question that has limited the interoperability the field hoped for.

Answer: A

A medical ontology and a manufacturing one may need incompatible treatments of parts, processes and substances, and forcing both into one scheme distorts at least one of them. Practical projects therefore tend to agree locally and translate between ontologies rather than unify. Category counts, expressiveness and testability are secondary. That a shared standard may be unattainable rather than merely unagreed is a conclusion worth reaching before a great deal of effort is spent.

Answer: B

Half a chair is not a chair, but half a quantity of butter is still butter, so the two require different treatment in an ontology. Substances are handled as stuff with properties that are intrinsic rather than extrinsic. Both are physical and both are representable, and countability is a linguistic reflection of the distinction rather than its basis. A distinction that language marks and that matters for inference is exactly the kind an ontology must capture explicitly.

Answer: C

Cutting a quantity of butter in half leaves both halves with the same density and melting point, which are intrinsic, but each has half the mass and volume, which are extrinsic. That division is what allows inference about parts of a substance. Internal structure, destructiveness and uniqueness are different notions. Classifying properties by whether they survive division is what lets an ontology reason correctly about quantities rather than only about individuals.

Answer: D

Reifying an event as an individual lets properties be attached to it, so one can say the meeting was long, was in London and involved three people, which a bare relation cannot express. That is why event calculus and similar formalisms treat events as objects. Transitions and functions are alternative treatments used in situation calculus and temporal logic respectively. Turning a relation into an object so that it can itself be described is the technique called reification and it recurs constantly.