Knowledge Representation Questions and Answers

Answer: C

Facts are the sentences put into the knowledge base directly, as distinct from the conclusions a reasoner derives from them. Whether they are true of the world is beyond the system's reach and is the responsibility of whoever supplied them. Verification and irrefutability are different notions. Keeping the asserted and the derived clearly apart is what allows a conclusion to be traced back to its premises when it turns out to be wrong.

Answer: D

The engine applies the rules of the representation's logic to derive conclusions, which is what distinguishes a knowledge base from a database. Validation, storage and presentation are supporting functions. The separation of what is known from how conclusions are drawn is the architectural idea the whole approach rests on, and it is what allows knowledge to be added without the reasoner being changed.

Answer: A

A database returns what was put in, while a knowledge base returns what follows, so a query may be answered by a chain of inference over facts none of which mentions the query directly. Format, query language and size are engineering differences of degree. That the answer to a question may be present only implicitly, and must be computed, is the whole point of the machinery and the reason its cost is worth paying.

Answer: B

A logical representation lets an agent believe a fact true, false or unknown, while a probabilistic one admits any degree of belief between, and that choice is separate from what the representation says exists. Designers' philosophies, required content and individual confidences are different notions. Distinguishing what a language says about the world from what it allows an agent to think about it is a useful axis on which to compare formalisms.

Answer: C

Propositional logic commits only to facts, first order logic to objects and relations, and temporal logic additionally to times, so each says something different about what there is. A particular ontology's categories are a use of that commitment rather than the commitment itself. Maintenance promises and completeness obligations are different matters. Comparing formalisms by what they assume the world is made of is what makes the choice between them a substantive rather than a stylistic one.

Answer: D

First order logic quantifies over objects and predicates their properties and relations, which is why it underlies almost every representation in the field. Propositional logic commits only to facts, probability theory adds degrees of belief to whatever ontology it is applied to, and temporal logic adds times to an existing commitment. Placing each formalism by what it assumes to exist organises them on a single scale rather than as unrelated alternatives.

Answer: A

Propositional logic treats each statement as an indivisible atom that is true or false, so it cannot express that something holds of every object of a kind. That limitation is why first order logic is needed for general knowledge, though propositional encodings remain practically important because their satisfiability problem is decidable. Times and degrees are added by other formalisms. Knowing what a language cannot say is what determines when it must be abandoned for a stronger one.

Answer: B

Modern satisfiability solvers handle millions of clauses, so encoding a bounded problem propositionally and solving it is often faster than reasoning in a richer logic. Hardware verification and planning both exploit this. Expressiveness is strictly less, semantics is required, and non-monotonic reasoning is not confined to any one logic. That a weaker formalism with extraordinarily good tools can outperform a stronger one is a consideration that pure expressiveness comparisons miss entirely.

Answer: C

Separating domain knowledge as data from a domain independent reasoner is what allows an expert to extend the system without programming, which was the founding promise of the knowledge based approach. Rewriting the reasoner, embedding rules in the base and attaching procedures to facts each erode that separation, the last being exactly the criticism of procedural attachment. An architecture whose central benefit is a separation must guard that separation carefully, since every convenience tends to erode it.

Answer: D

The field supplied reification, defaults, description logics, the open and closed world distinction and the web ontology standards, all of which remain in use. No universal representation emerged, learned models plainly do represent knowledge in their own way, and no complete commonsense ontology exists. Judging a research programme by the tools and concepts it leaves behind, rather than by whether its original goal was met, is usually the fairer assessment.