donovan@csl.sri.com (Donovan Hsieh) (01/25/91)
Six months ago, I posted a request for input and information regarding the
current research progress on the tight coupling of knowledge-based systems
with database systems. Since the, I received one response from Frank
Lebastard,
the developer of DRIVER system at INRIA, France. Now a full technical report
has been completed on this topic. The abstract is shown in the following.
If any one should be interested in this report, you can either email me or
contact SRI International for the CSL Technical Report # SRI-CSL-90-15.
Donovan Hsieh
M/S: EL239
SRI International
Computer Science Lab
333 Ravenswood Ave.
Menlo Park, CA 94025
email: donovan@csl.sri.com
phone: (415)859-4123
*****************************************************************************
A Logic to Unify Semantic Network Knowledge
Systems with Object-Oriented Database Models
Donovan Hsieh
Computer Science Lab
SRI International
In recent years, coupling knowledge-based systems with existing
database systems has gained tremendous attention in both artificial
intelligence and database communities. Historically, there have
been various ways to encode knowledge, such as rule-based,
frame-based, and logic-based knowledge representations. In this report,
we examine some approaches that have been proposed to loosely
couple knowledge systems with predominantly relational database systems.
Many existing approaches suffer from (1) poor support of the
semantic modeling capabilities in the underlying relational database
systems that is required by the coupled knowledge systems; (2)
inadequate performance due to the separation of inference engines
from database engines; (3) weak knowledge consistency due to
the separation of data and knowledge stored in different domains.
In our Persistent Data/Knowledge Base System, we take a novel approach
by tightly coupling semantic-network based knowledge systems with
object-oriented database models. In order to support the highly
semantic knowledge-based system, we need to augment existing
object-oriented data models with extra modeling capabilities, such as
the interaction of generalization/specialization with aggregation
and other association relationships in semantic nets.
We also argue that semantic-network based knowledge systems provide
a natural way to integrate both data and knowledge in a unified
framework when they are stored persistently in object-oriented
databases with extended modeling capabilities. We also illustrate
the similarity between semantic-network knowledge representation
and those semantic database models that, from the modeling power
perspective, can be considered as super set of object-oriented
data models
Finally, we describe future research issues which include
incorporating logic representation with high order predicate
languages to encode other knowledge, such as uncertainty,
database constraints, and object class/method definitions.
**************************************************************************