Professional Documents
Culture Documents
Approaches in
7 Knowledge
Management
LEARNING OUTCOMES
By the end of this topic, you should be able to:
1. Identify the components of knowledge management systems;
2. Identify the role of artificial intelligence in managing knowledge;
3. Recognise different types of artificial intelligence tools used for
knowledge management;
4. Discuss the implementation of knowledge management systems; and
5. Describe how knowledge engineering processes are used in
developing knowledge-based systems.
X INTRODUCTION
Because we do not understand the brain very well we are constantly tempted to
use the latest technology as a model for trying to understand it. In my childhood
we were always assured that the brain was a telephone switchboard. („What else
could it be?‰) I was amused to see that Sherrington, the great British neuroscientist,
thought that the brain worked like a telegraph system. Freud often compared the
brain to hydraulic and electro-magnetic systems. Leibniz compared it to a mill, and
I am told some of the ancient Greeks thought the brain functions like a catapult. At
present, obviously, the metaphor is the digital computer.
The quotation shows the important role of technology for humans, and how our
perception of technology changes as it evolves over time.
However, not all types of knowledge can be managed successfully through the
use of technology as some are better managed through human-oriented
processes with the support of ICT. Tsui (2005) believes that successful
implementation of any KM project involves the blending of technology, people,
process and content. To select the appropriate technology support for KM
requires an understanding of the extent to which knowledge can be structured
and the type of strategy adopted: codification versus personalisation strategy.
The codification strategy relies on knowledge which is stored in databases that
are easily accessible by people who need to access it. The personalisation
strategy, on the other hand, focuses on the tacit dimension of knowledge that is
embedded in people and is shared through person-to-person contact.
ACTIVITY 7.1
Storage and retrieval technologies are originally meant for using a database
management system to store and manage knowledge. This system worked well
in the early days for storing and managing explicit knowledge, and tacit
knowledge that are converted to explicit knowledge. Nevertheless, capturing,
storing and managing tacit knowledge requires a different set of tools. Electronic
document management systems and specialised storage systems that are part of
collaborative computing systems fill this void. These storage systems are
commonly known as knowledge repositories or organisational memory.
Table 7.1 shows the relationship between these knowledge management systems
and the Web.
TOPIC 7 TECHNOLOGICAL APPROACHES IN KNOWLEDGE MANAGEMENT W 135
Knowledge
Web Impacts Impacts on the Web
Management
Communication Consistent, friendly, graphical user Knowledge captured and
interface for client units shared is utilised in
improving
Improved communication tools
communication,
Convenient, fast access to knowledge communication
and knowledgeable individuals management and
communication
Direct access to knowledge on the technologies
servers
Collaboration Improved collaboration tools Knowledge captured and
shared is utilised in
Enables anywhere/anytime
improving collaboration,
collaboration
collaboration
Enables collaboration between management, and
organisation, vendors and customers collaboration technologies
(Group Support Systems)
Enables document sharing
Improved, fast collaboration and links
to knowledge sources
Makes audio and video conferencing
a reality, especially for individuals
not using a Local Area Network
(LAN)
Storage and Consistent, friendly, graphical user Knowledge captured and
Retrieval interface for clients shared is utilised in
improving data storage
Server provides for efficient and
and retrieval systems,
effective storage and retrieval of
database
knowledge
management/knowledge
repository management,
and database and
knowledge repository
technologies
136 X TOPIC 7 TECHNOLOGICAL APPROACHES IN KNOWLEDGE MANAGEMENT
SELF-CHECK 7.1
In the early 1970s and 1980s, AI development was ready for the technology that
KM would bring. Computer scientists from the AI field strongly believed that
expert knowledge could be codified, directly transferred and managed through
an expert system. However, this proved to be a costly mistake as human
knowledge was much more complex and context dependent than was first
thought; in the end, not all of it could be coded into a computer program.
ACTIVITY 7.2
systems are two well-known examples of early and successful expert systems.
This has led to the birth of knowledge engineering, a domain that supports the
development of these systems.
Expert systems continue to evolve as the need to have a stable technology for
managing knowledge grows and their current role as an enabler for KM
initiatives has led to greater appreciation of this technology. As a result of this
evolutionary process, different names have been given to this technology to
reflect its current impact and adoption as an established tool for managing
knowledge, business rules and process automation in software systems. In recent
years, the terms knowledge-based systems (KBS), business rule management
systems (BRMS), rule-based systems, and knowledge systems (KS) have been
used interchangeably with the term expert system. They all refer to the same type
of system, where knowledge (in the form of rules) is inferred in order to arrive at
a decision.
To comprehend this situation better, let us look at the technique how fuzzy logic
would represent various temperatures in a computer application to control room
temperature automatically. These terms (known as membership functions) are
imprecisely defined so that, for example in Figure 7.1, cold is between 50 degrees
and 70 degrees (in Fahrenheit), although the temperature is most clearly cool
from about 60 to 67 degrees. Note that the values of cold, cool and warm can
overlap with each other. To control the room environment using this logic, the
programmer would develop similarly imprecise definitions for humidity and
other factors, such as outdoor wind and temperature. One of the rules would be:
„If the temperature is cool or cold and the humidity is low, while the outdoor
wind is high and the outdoor temperature is low, raise the heat and humidity
in the room‰. The computer would then combine the membership function
readings in a weighted manner, and using all the rules, raise and lower the
temperature and humidity.
TOPIC 7 TECHNOLOGICAL APPROACHES IN KNOWLEDGE MANAGEMENT W 141
Figure 7.1: Membership functions for the input temperature in the logic of the thermostat
to control room temperature
Fuzzy logic systems are also useful in management decision making and
organisational control. It is used to detect possible fraud in medical claims, help
stock traders select companies for potential acquisition and decide risk categories
for insurance applications.
A neural network contains a large number of sensing and processing nodes that
continuously interact with each other. Figure 7.2 shows one type of neural
network consisting of an input layer, an output layer and a hidden processing
layer. People train the network by providing it with a set of training data for
which the input produce a known set of outcomes or conclusions. This enables
the computer to learn the correct example. As more data are entered into the
computer, each case is compared with the known outcome. If the outcome is
different from the expected one, then a correction is calculated and applied to the
nodes in the hidden processing layers. These steps are repeated until a specific
condition is met, for example, correction of less than 5% is achieved. The neural
network shown in Figure 7.2 has learned how to identify a fraudulent credit
purchase.
Although neural networks are very useful, there are several puzzling aspects
about them. Unlike a knowledge-based system, which has the ability to explain
why it has arrived at a solution, neural networks cannot always explain how a
solution is reached. Furthermore, neural networks cannot always guarantee a
completely certain solution, arrive at the same solution again with the same input
data, or guarantee the best solution. They are also very sensitive and may not
perform well if their training is too little or too much. Currently, neural networks
are used to aid human decision makers.
There exist many intelligent agent applications nowadays and these can be found
in operating systems, application software, e-mail systems, mobile computing
software, virus detection programs and network tools. For example, wizards
found in Microsoft Office software tools have built-in capabilities to show users
how to accomplish various tasks such as formatting documents, or drawing
tables, and to anticipate when a user requires assistance. Several names are used
to describe intelligent agents and these include software agents, wizards,
software daemons, knowbots, softbots and bots (intelligent software robots).
These terms sometimes refer to agent of different types or intelligence levels.
Procter & Gamble (P&G), manufacturer of personal care, food and household
products, used agent-based modelling to improve coordination between different
members of its supply chain in response to changing business environments.
Figure 7.3 illustrates the use of intelligent agents in Procter & GambleÊs supply
chain network. The network models a complex supply chain as a group of
semiautonomous agents representing individual supply chain components such
as trucks, productions facilities, distributors and retailers. The behaviour of each
agent is programmed to follow rules that mimic actual behaviour, such as
144 X TOPIC 7 TECHNOLOGICAL APPROACHES IN KNOWLEDGE MANAGEMENT
„dispatch a truck when it is full‰. Using agents, the company can do simulations
to perform what-if analyses on inventory levels, ion-store stockouts and
transportation costs.
Using intelligent agent models, P&G found that trucks must be dispatched before
they are fully loaded. Although transportation costs would be higher for partially
loaded trucks due to driver time and fuel to deliver fewer goods, the simulation
showed that retail store stockouts would happen less frequently, thus reducing
the amount of lost sales, which would be more than the higher distribution cost.
Through agent-based modelling, P&G saved about $300 million annually on an
investment that is around 1% of the saving amount.
ACTIVITY 7.3
1. Briefly discuss how knowledge expert systems have evolved.
2. What is case-based reasoning and how is it used in managing
knowledge?
3. Argue the importance of having fuzzy values compared to precise
values.
4. Describe the limitations of neural networks compared to
knowledge-based systems.
5. Justify why intelligent agent technology is useful for knowledge
management.
The greatest challenge of KMS is to identify and integrate the three components
of communications technologies, collaboration technologies and retrieval
technologies, to meet the knowledge management needs of the organisation.
Earlier knowledge management systems were built with networked technology –
intranets, collaborative computing tools – groupware and databases – knowledge
repository. These systems were constructed from a variety of off-the-shelf
IT components. Larger management consulting firms like Accenture and
J. D. Edwards developed their own knowledge architecture with a set of tools
that combined all the three types of technology. Collaborative computing suites
like IBMÊs Lotus Notes/Domino and GroupSystemsÊ OnLine also provide many
KMS capabilities. Other systems integrate different sets of tools from a single or
multiple vendors, like J. D. EdwardsÊ Knowledge Garden KMS which integrates
Microsoft tools and products. However, the KMS technology has evolved to
integrate the three components into a single package.
146 X TOPIC 7 TECHNOLOGICAL APPROACHES IN KNOWLEDGE MANAGEMENT
The earlier enterprise information portals in the market did not have
knowledge management features, however, now most of them do.
Leading portal vendors include Autonomy, Brio, Corechange,
Dataware, Intraspect, Hummingbird, InXight, IBM/Lotus, Knowmadic,
OpenText, Plumtree, Verity, Viador and Vignette. Database vendors
such as Oracle, Sybase and Microsoft are also selling knowledge
portals. Portal prices typically range from a few hundred thousand
to a few millions depending on the requirements and the size of the
organisation. One popular knowledge portal in Malaysia is
myGoverment – the Malaysian governmentÊs official portal.
SELF-CHECK 7.2
ACTIVITY 7.4
Early versions of KBS were built around expert knowledge, as KE activities were
approached as a transfer process; however, this approach lacks the problem-
solving capabilities of the expert. Nevertheless, KBS developers quickly
discovered that such capabilities could only be captured through the use of
conceptual models in order to understand the problem-solving behaviour of the
expert. This leads to defining KE as a modelling process. Sections 7.6.2 and 7.6.3
will discuss this in more detail.
KBS testing is done in two phases: verification and validation of the system. In
the verification phase, the rules in the knowledge base are analysed for sequence,
structure and specification to ensure the logical correctness of the rules. Then, the
validation of the KBS is carried out to test the behaviour of the system in a
realistic situation. There are well-established techniques for the verification and
validation of KBS which are dependent on the implementation domain of the
system. For example, in safety-critical applications such as aeroplanes and space
missions, the reliability of the KBS is essential, and therefore a formal method
verification is essential, whereas in a low-risk application such verification is not
necessary. However, testing can also be done on the correctness of the rules
during the iterative development process.
The knowledge acquired from the expert is logically checked for its correctness
before populating the knowledge base. Reliability of the knowledge base is
achieved by removing circular rules that are contradictory in meaning or logic,
deleting redundant rules that provide different methods for the same problem
which causes knowledge duplication, and removing unusable rules that never
execute because of the contradictions in the premise of the rule.
TOPIC 7 TECHNOLOGICAL APPROACHES IN KNOWLEDGE MANAGEMENT W 155
Knowledge
Acquisition
Coding Coding
Testing Testing
Implementation Implementation
Maintenance Maintenance
In comparison with SE, the KE has one additional stage: that of knowledge
acquisition (KA). This stage is vital in KBS development as the KBS is designed
around the domain expertÊs knowledge of solving problems for a particular task,
such as diagnosis, assessment and so on. The acquired knowledge is then used to
populate the knowledge base in the form of rules, with which the system will
perform reasoning. However, in SE there is no KA stage as the system is
intended to capture information rather than reason with it and the actual dataset
of the database will be populated by the system user when the system is
deployed. Therefore, it may be concluded that the KA stage differentiates the SE
and KE domains when developing software systems.
156 X TOPIC 7 TECHNOLOGICAL APPROACHES IN KNOWLEDGE MANAGEMENT
The transfer approach also ignores the importance of the tacit knowledge
of an expertÊs problem-solving capabilities. This creates a new problem if the
knowledge base is to be updated, as changes require substantial effort in
reconstituting the coded rules in order to implement the needed changes.
Consequently, the transfer approach is only feasible for developing prototype
systems and fails to scale up when building larger and more reliable KBSs where
knowledge bases change. These deficiencies have caused the transfer approach to
be replaced by the modelling approach. During this time, the SE community had
already used the modelling approach to construct information systems and it
seems to also suit KBS development.
Another direction taken by the KE community during this time to overcome the
limitations of the knowledge transfer approach is through Knowledge Sharing
initiatives and the major outcomes of this work are ontologies, knowledge
interchange format (KIF), Knowledge Query and Manipulation Language
(KQML) and Knowledge Representation System Specification.
Conceptual Model
If income = 3000
then credit = 5500
If income = 4000
then credit = 7500
Inference Engine
Expertise Implemented System
Developers feel that building a KBS means building a computer model that has
problem-solving capabilities similar to those of a domain expert. It is not
necessary to be an exact replica of human cognition; instead, it must simulate the
158 X TOPIC 7 TECHNOLOGICAL APPROACHES IN KNOWLEDGE MANAGEMENT
thinking processes involved in the narrow area of concern. While experts may
consciously articulate part of their knowledge, they will not be aware of a
significant part of this knowledge because it is hidden in their skills. This view
has been an important part of KE activities.
SELF-CHECK 7.3
the reasoning process of the inference engine to generate decisions related to the
problem domain. These features are unique to KBS and as such are commonly
used to define KBS in the literature.
7.7.1 Architecture
Architecture differentiates a knowledge-based system from an information
system. The reasoning engine (inference engine) and the knowledge base are the
main constituents of KBS architecture. This basic architecture was originally
developed and used in expert systems in the late 1970s and is still in use today.
The inference engine is usually programmed in a shell-based programming
language rather than developed and run with explicit declarative knowledge and
information to arrive at a conclusion. The knowledge base contains all the
domain knowledge represented as rules (production rules) that are to be
consumed by the inference engine during execution.
The current use of this architecture is a modified version of the original one. The
original architecture is shown in Figure 7.7. Here, the reasoning control actually
refers to the reasoning or inference engine, and application domain knowledge
refers to the knowledge base of the domain.
Reasoning control
Figure 7.7: The basic architecture of the first generation of expert systems
TOPIC 7 TECHNOLOGICAL APPROACHES IN KNOWLEDGE MANAGEMENT W 161
Inference
Explanation facility
Mechanism
Knowledge
Expert Knowledge base
acquisition facility
Based on the KBS architecture presented in Figures 7.7 and 7.8, it may be
concluded that the core components of any given KBS are the knowledge
base and the inference engine (or reasoning mechanism). However, current
architectures differ from the original because new components such as
knowledge acquisition, user interface, and the explanation facilities are now
added to the core components to make the architecture much more suitable for
those current practices used in systems development. One example of this is the
162 X TOPIC 7 TECHNOLOGICAL APPROACHES IN KNOWLEDGE MANAGEMENT
user interface component, which has become an important part of any systems
development project and takes the form of a conventional graphical user
interface (GUI). Due to the growing importance of the user interface element in
any system, this component has been explicitly included in the current KBS
architecture.
Most knowledge systems adopt rules to drive their inference engines. Earlier
inference engines (such as CLIPS, VP-Expert, XeprtRule and KnowledgePro)
used shell-based production rule systems. These were written in a declarative
rather than procedural programming style based on algorithms such as RETE.
However, there have been developments in inference engines in which support
for embedding features in conventional programming languages such as C++
and Java are implemented, which simplifies the integration of conventional
program code with rule inferencing capabilities.
As a result, the Java Expert System Shell (Jess), based on C-Language Integrated
Production System (CLIPS), has been developed to enable enterprise software
developed using Java to have some built-in reasoning capabilities. Use of the Java
programming language to develop rule-based applications has prompted the
Java community to develop standards for Java-based rule engines based on the
JSR-94 Java Rule Engine API. The JSR-94 specification is popular among vendors
and is implemented in ILOG JRules, Jess, Fair Isaac Blaze Advisor, Computer
Associates CleverPath Aion, Drools and others.
The KBS technology has evolved from the early rule-based reasoning to
accommodate other strands of AI research, such as fuzzy logic, genetic
algorithms, case-based reasoning and neural networks. This evolution has been
beneficial to the knowledge management initiatives community as different KBS
technologies can be utilised in providing solutions to the problem domains.
Nevertheless, production rules are considered as the most convenient approach
in representing most business rules and are widely supported by many inference
engines. Section 7.7.3 discusses the benefits and problems of using KBS as a tool
for managing knowledge.
of the KBS to produce some decision for all possible inputs, while consistency is
the KBSÊ ability to produce a standard set of decisions that are true for all
possible inputs. Rule verification identifies redundant rules, inconsistent rules,
circular rules and unreachable decisions. Validation of the KBS is done by
executing the system and comparing the test results against the required
performance. This proves that the KBS is producing decisions only for the set of
given inputs. Validation and verification is an important area in KE and any KBS
that is crucial to safety and health decisions must be verified and validated; this
contrasts with those systems that are not safety or mission critical.
Using KBS the quality of the decisions made increases because there are fewer
inconsistencies than if the decisions were performed manually. Results produced
by the KBS are consistent throughout its operational lifespan unless it is modified
to incorporate new rules or delete older ones. Two copies of the same KBS will
provide the same answer to the same problem; human experts do not achieve
this level of consistency and such consistency is important in certain domains
such as insurance premium calculations for insurance policies. Achieving such
consistency is vital as decision quality is an important criterion when adopting
KBS, particularly in relation to decisions involving huge amounts of data,
variables and information.
KBSs are also capable of assisting experts in decision making even if the experts
have that knowledge to hand; this improves the accuracy and timeliness of
decisions made. Experts are humans, who have the tendency to forget and make
mistakes when making decisions. However, when the knowledge of the experts
is stored as rules in the knowledge base, such mistakes can be avoided provided
there are no implementation errors. KBSs will always produce the desired result
for every decision case, as they will not leave out any rule (consideration) in the
reasoning process. The decision made will always be the same and reliable.
Knowledge Knowledge
Acquisition Acquisition
Coding
Maintenance
ACTIVITY 7.5
1. Describe the basic components in a KBS architecture, and the
additional components in the current KBS architecture.
2. Discuss three important benefits of using KBS in managing
organisational knowledge.
3. Argue why current KBS are usually embedded or integrated into
other larger applications.
4. Compare the activities in knowledge engineering and KBS
development, and justify why conceptual modelling in KBS
development is important.
ACTIVITY 7.6
1. Discuss the five artificial intelligence tools that are widely used
for knowledge management.
2. Provide a brief description for the following terms:
(a) Knowledge engineering
(b) Knowledge-based systems
(c) Electronic Document Management
3. Justify why the modelling process is better than the transfer
process in knowledge engineering.
4. Argue the benefits of knowledge-based systems as a tool in
managing organisational knowledge.
5. Explain why the role of artificial intelligence in knowledge
management is controversial.
TOPIC 7 TECHNOLOGICAL APPROACHES IN KNOWLEDGE MANAGEMENT W 169
x Not all types of knowledge can be managed successfully through the use of
technology as some types of knowledge are better managed through human-
oriented processes with the support of ICT.
x AI methods and tools are widely used to assist in finding expertise, eliciting
knowledge automatically and semi-automatically, interfacing through natural
language processing, and intelligent search through intelligent agents. AI
methods such as knowledge-based systems, neural networks, fuzzy logic,
evolutionary algorithms and intelligent agents, are used in knowledge
management systems.
Awad, E., & Ghaziri, H. H. (2004). Knowledge management. New Jersey: Prentice
Hall.
Becerra-Fernandez, I., Gonzalez, A., & Sabherwal, R. (2004). Knowledge
management – Challenges, solutions, and technologies. New Jersey:
Prentice Hall.
Debowski, S. (2006). Knowledge management. New Jersey: John Wiley & Sons.
Loudon, K. C., & Laudon, J. P. (2007). Management information systems –
Managing the digital firms. New Jersey: Prentice Hall.
Natarajan, S., & Shekhar, S. (2000). Knowledge management: Enabling business
growth. New Delhi: McGraw-Hill.
Turban, E., Leidner, D., McLean, E., & Wetherbe, J. (2007). Information
technology for management: Transforming organisations in the digital
economy. Boston: John Wiley & Sons.