Databases Reference
In-Depth Information
Databases
Expert systems
Management of big volumes of data
Reasoning capabilities
Extensional data: facts
Integrity, recovery, query optimization
Maintained by administrators
Intensional knowledge: rules
Knowledge representation
Maintained by experts
Require an efficiency improvement to
manage big volumes of information
Need of reasoning capabilities
internal to the database
Deductive databases?
Figure 4.1
Parallel evolution of DBs and expert systems.
Data are represented by means of extensions of DB predicates (i.e.,
facts), while knowledge is represented by the intension of DB predicates.
Knowledge, or intensional information, is defined by means of rules that
allow us to deduce new information from that explicitly stored in the DB.
Two kinds of rules are usually considered: deductive rules, which allow us to
define new facts (view or derived facts) from stored facts, and integrity con-
straints, which state conditions to be satisfied by the DB.
Deductive DBs are able to manage most of the current applications,
being specially suited for those applications that are characterized by the need
to perform more complex manipulations on the data rather than just query-
ing or updating base facts [1]. Deductive rules and integrity constraints,
together with reasoning capabilities, ease the sharing of common knowledge
within more complex application domains, thus facilitating knowledge
reusability.
Deductive DBs also can be understood as the result of the application
of logic (and artificial intelligence (AI) aspects) in the traditional DB field
since they use Datalog, a language derived from logic programming, as a rep-
resentation language, to define the contents and the structure of the informa-
tion stored in the DB. Datalog is a declarative, nonprocedural language,
uniform and set oriented. The impact of logic in deductive DBs is widely dis-
cussed in [2, 3].
From a historical perspective, deductive DBs can be seen as extensions
of the relational ones as well, mainly because the expressive power of their
data definition language is greater than that of relational DBs. This impor-
tant issue is not so clear at the present moment and will deserve further dis-
cussion in Section 4.2.
