184 8 Query Languages
- A language based on rule engines. As we noted in section 3.1, a query is
just a rule with no conclusion, just a condition. Thus any rule language
provides, as a special case, a query language. However, the same problem
arises here as in the possibility above. Rules engines generally do not
support the logic used in the Semantic Web languages. - A language based on formal logic. Processing a language such as OWL
requires a theorem prover in general. As discussed in section 3.3, one
queries a theorem prover by presenting it with a conjecture. This is not
the same as a query, because a theorem prover will only report whether
the query can be satisfied or not.
None of these choices is completely satisfactory. In two of the cases, one
would be ignoring the semantics of RDF and OWL, while in the third case,
the query language would be unsatisfactory for nearly all uses. To under-
stand the problem, suppose that one would like to know the number of
atoms in the nitrous oxide molecule as in exercise 8.2. This is such a simple
and obvious kind of query that it seems amazing that any query language
would have any trouble with it. However, in the Semantic Web languages,
the only statement that is entailed by the available information is that there
areat leasttwo atoms. The fact that there areexactlytwo atoms is not en-
tailed. The reason for this has to do with the monotonicity of the Semantic
Web languages, as explained in section 4.4. A statement is entailed when it is
true in every model of the theory. Unfortunately, there are models for the ni-
trous oxide theory that have more than two atoms. One could certainly add
additional statements to the nitrous oxide theory which would allow one to
entail the fact that there are only two atoms in the molecule. However, that
would only resolve this particular query. It is not feasible to add all of the
additional statements to a knowledge base that would be needed to resolve
all possible queries. In fact, one can prove that it is logically impossible to
do this. As a result, there can never be a fully satisfactory query language
that will preserve the logic and semantics of the OWL languages as they are
currently defined. The same is true for the transformation tasks which are
discussed in Chapters 10 and 11.8.4 Exercises
- Using the health study database in section 1.2, find all interviews in the
year 2000 for which the study subject had a BMI greater than 30.