Databases Reference
In-Depth Information
does not adopt the
unique name assumption
(UNA, for short) according to which, for any interpretation
OWL 2
Unlike standard DLs,
I
=(
Δ
I
,
·
I
)and
any distinct individual names
a
i
and
a
j
,wemusthave
a
i
=
a
j
.Thatiswhy
OWL 2 QL
=
a
j
in its syntax.
Theorem 1 ([11,4]).
(
i
)
The satisfiability problem for
has inequality constraints
a
i
OWL 2 QL
knowledge
0
for data
bases is
NLogSpace
-complete for combined complexity and in
AC
complexity.
(
ii
)
CQs and
OWL 2 QL
TBoxes are FO-rewritable, and so CQ answering
0
for data complexity.
over
OWL 2 QL
TBoxes is in
AC
We will explain how one can compute FO-rewritings for CQs over
OWL 2 QL
TBoxes in the next section. Meanwhile, we are going to illustrate the expressive
power of the language and discuss whether it can be extended without losing
FO-rewritability. The following example shows what can and what cannot be
represented in
is to facili-
tate OBDA with relational databases, our example is from the area of conceptual
data modelling.
OWL 2 QL
TBoxes. As the primary aim of
OWL 2 QL
Example 5.
Consider the UML class diagram in the picture above representing
(part of) a company information system. According to the diagram, all managers
are employees and are partitioned into area managers and top managers. In
OWL 2 QL
, we can write
Manager
Employee,
AreaManager
Manager,
TopManager
Manager,
AreaManager
TopManager
⊥
.
However, the covering constraint
Manager
AreaManager
TopManager
uses
union
. Each employee has two attributes,
empCode
and
salary
, with integer values. Unlike OWL, here we do not distinguish
between abstract objects and data values (there are, however, approaches based
on concrete domains [5,64]). Hence we model a datatype, such as
Integer
,bya
concept, and an attribute, such as employee's salary, by a role:
, which is not allowed in
OWL 2 QL
salary
−
Employee
∃
salary,
∃
Integer.