Databases Reference
In-Depth Information
G
K
and returns
'false' if this match is spurious (that is, cannot be realised in the real canonical
model
RDBMS. This procedure takes as input a match of the query in
C
K
) and 'true' otherwise. The filtering procedure runs in polynomial time,
although it may have to run exponentially many times (to check exponentially
many matches for the same answer tuple) in the worst case.
6OBDAw th
Ontop
In this final section, we consider the architecture of the OBDA system
Ontop
and available as a plugin for the ontology editor Protege 4, a SPARQL end-
point and OWLAPI and Sesame libraries.
In OBDA with databases, the data comes from a relational database rather
than an ABox. From a logical point of view, a
database schema
[1] contains
predicate symbols (with their arity) for both stored database relations (also
known as tables) and views (with their definitions in terms of stored relations)
as well as a set
Σ
of
integrity constraints
(in the form of functional and inclusion
dependencies; for example, primary and foreign keys). Any instance
I
of the
database schema must satisfy its integrity constraints
Σ
.
The vocabularies of the database schema and the given TBox are linked
together by means of mappings produced by a domain expert or extracted
(semi)automatically. There are different known types of mappings: LAV (local-
as-views), GAV (global-as-views), GLAV, etc.; consult, e.g., [41] for an overview.
Here we concentrate on GAV mappings because they guarantee low complexity
of query answering (in what follows we call them simply mappings). A
mapping
,
M
, is a set of rules of the form
S
(
x
)
←
ϕ
(
x
,
z
)
,
where
S
is a concept or role name in the ontology and
ϕ
(
) a conjunction
of atoms with database relations (both stored and views) and a
filter
,thatis,
a Boolean combination of built-in predicates such as = and
<
. (Note that, by
including views in the schema, we can express any SQL query in mappings, which
is important from the practical point of view.)
Given a mapping
x
,
z
M
OWL 2 QL
T
from a database schema to an
TBox
and
I
an instance
of this schema, the ground atoms
S
(
a
)
,
for
S
(
x
)
←
ϕ
(
x
,
z
)in
M
and
I
|
=
∃
z
ϕ
(
a
,
z
)
,
comprise the ABox,
[59]
(this ABox is just a convenient presentational tool and does not have to be
materialised by the system). We can now define
certain answers
to a CQ
A
I
,M
, which is called the
virtual ABox
for
M
over
I
q
over
T
A
I
,M
).
As an illustration, we consider a (simplified) database IMDb (
www.imdb.com/
interfaces
), whose schema contains relations
title
[
m, t, y
] with information
about movies (ID, title, production year), and
castinfo
[
p,m,r
] with information
linked by
M
to
I
as certain answers to
q
over (
T
,
