TRENDS IN DATABASE TECHNOLOGY - Database Design and Development: An Essential Guide for IT Professionals - page 665

Deductive Databases
Interest in deductive databases began in research establishments in the late 1970s
in an effort to combine logic and database technology. During that period, the
promises of artificial intelligence and expert systems attracted a lot of attention.
Research projects abounded in Europe, Japan, and the U.S. Japan launched the Fifth
Generation Project to encompass a number of technologies including logic systems
and database technology. Research institutions proposed several experimental
deductive database systems. A few commercial systems were also deployed.
Major objectives that propelled the effort included:
• Extend the relational model and develop a new type of database system.
• Improve the functionality of such a database system to perform as a deductive
DBMS.
• In addition, enable the new type of DBMS to support general-purpose
applications
Deductive databases combine the functionality of logic programming, methods
to specify rules about data, and an inference engine. A declarative language
expresses the rules. Logic similar to that used in Prolog language is applied. The
inference engine performs as a deduction tool to deduce new facts from the data-
base. Datalog, a variation of Prolog, defines the rules as they relate to relations in
the database.
In a deductive database, facts and rules are specified.
Facts. Defined similar to relations in a relational database, except that attribute
names are not included. A relation represents a real-world object, and its attributes
attempt to describe the object, perhaps only partially. In a deductive database, the
position of an attribute within the tuple determines what the value of an attribute
means.
Rules. Views in a relational database system define virtual relations, not what
is actually stored. In a sense, rules are like views. They specify virtual units that
can be derived from the facts by applying inference techniques. Defined similar
to data views in a relational database, except that attribute names are not
included. Views cannot include recursive definitions; however, rules may involve
recursion.
Backward chaining and forward chaining methods are used in deductive data-
base systems to evaluate query or transaction results.
Figure 20-25 presents Prolog notation to express facts, rules, and queries. This is
not an explanation of how queries work in the deductive database system—just a
sample to indicate the notation.
Multimedia Databases
Traditional database systems store and manage textual and numeric data. The intro-
duction of object-relational database systems meant a big leap forward by the inclu-
sion of new data types. With UDTs, users can define new data types. Still this is not