CHAPTER 3

Overview of Expert Systems

• Introduction
• History of expert systems
• Architecture of an expert system
• Applications of expert systems
• Application of expert systems in LIS work
• Applications of expert systems in cataloguing
• References
Chapter 3: Overview of Expert Systems 38

3.1 Introduction
Expert systems are a part of Artificial Intelligence (AI). Artificial Intelligence is
that field concerned with the computations that connect situations to complex,
human-like actions'. AI can help or to some extent replace human experts in
problem-solving. Though much of the research in AI started in early 20* century,
it was the British mathematician Alan Turing who brought AI into the centre
stage of computer science. He is considered as the father of AI. Turing is better
known for his "Turing Tesf experiment which was described in his classical
paper "Computing machinery and intelligence" ^vi!o\\s\itd in 1950.

The term Artificial Intelligence was coined by John McCarthy during a

Conference held at Dartmouth College in Hanover, New Hampshire in 1956. Two
years later McCarthy invented the programming language LISP for developing AI
applications.

In 1967, Richard Greenblatt at MIT built a knowledge-based chess-playing

program, MacHack, that was good enough to achieve a class C rating in
tournament play.

AI is concerned with developing computer systems which reason and behave

in ways a neutral observer would consider "human". The areas in AI include:
Problem solving, Knowledge Based Systems, Learning, Natural Language
Processing, Vision, Speech, and Game playing, etc^.

Some branches of AI have moved beyond pure research. They seek to apply
what is known as AI science; they produce marketable AI technologies. One
such branch is robotics, which seeks to give computers human mobility and
sensory capabilities. Another is, knowledge based systems.
Chapter 3: Overview of Expert Systems 39

Winston'^ gives some of the key AI areas:

• Computational robotics, which includes vision, manipulation,

locomotion, reasoning about space, and reasoning about object interaction

• Sensor interpretation, which includes visual sensors, acoustic sensors,

and signal-to-symbol conversation

• Expert systems, which includes rule-based reasoning, constraint-based

reasoning and analogy-based reasoning

• Language understanding, which includes speech, syntactic analysis,

question understanding and discourse understanding

• Learning, which includes learning by noting patterns, and learning by

exploiting precedents

• Common-sense reasoning, which includes variable-precision reasoning

and qualitative reasoning about quantities

• Supercomputing hardware and software, which includes massively

parallel architecture and very high-level languages

The present work deals with Expert Systems (ES) only and hence it is explained
in detail.

Expert systems are considered to be the most important contribution of AI. Terms
such as "expert systems," "knowledge-based systems," and "decision support
systems" are used more or less synonymously by some writers, while others
Chapter 3: Overview oj Expert Systems *i, ''*!'. 40

emphasize the subtle differences between such systems^. An expert system may
be defined as a computer program that exhibits, within a specific domain, a
degree of expertise in problem solving that is comparable to that of a human
expert^. Expert systems are meant to support the domain experts in quick
decision making and thereby problem solving.

Expert systems are computer systems designed to simulate the behaviour of

human experts in a narrow domain of knowledge. They are relatively new and
growing branch of AI. Expert systems are based on the idea that humans
commonly use many types of knowledge when solving problems. Knowledge
is an integrated collection of information which, when used, produces a
competent level of performance. Two of the most important forms of knowledge
are factual knowledge and heuristic knowledge.

The factual knov/ledge is the type of learning a person usually gains from reading
books and attending school. Factual knowledge provides a set of general theories.
It can suggest highly structured solution processes called algorithms, for solving
problems in a finite number of steps.

Heuristic knowledge is the type of learning a person gains from experience.

Heuristics are rules of thumb or situation-specific theories. Since each person's
experiences are unique, heuristics knowledge is personal and private. People use
heuristic knowledge every day to solve problems.

Expert systems differ in important ways from both conventional data processing
systems and systems developed in other branches of AI. Firstly, they perform
different tasks at expert levels of performance. Secondly, they emphasize domain-
specific problem-solving strategies over the more general "weaker methods" of
AI. Thirdly they employ self knowledge to reason about their own inference
Chapter 3: Ovei-view of Expert Systems 41

processes and provide explanations for conclusions reached^. In other words, an

expert system can be compared with a computer based system that achieves a
high level of performance in the task areas for which humans require years of
special education and training.

The usefulness of an expert system is best explained by Waterman^ in the

following paragraph:

The most useful feature of an expert system is the high-level expertise it

provides to aid in problem solving. This expertise can represent the best
thinking of the top experts in the field, leading to problem solutions that
are imaginative, accurate, and efficient. It's the high-level expertise
together with skill at applying it that makes the system cost-effective, able
to earn its own way in the commercial market place.

Ignizio^ credits experts systems with following advantages over human experts:

• Expert systems are always and instantly available, and always perform at
the same level of expertise.

• Expert system has direct and instantaneous access to the necessary

databases (ideally) and is not bound by limited, biased, and imperfect
recollections of the human.

• It is logical, objective, and consistent - and thus unswayed by emotional

arguments that might influence a human

• Expert systems do not forget or make mathematical errors

Chapter 3: Overview uf Expert Systems 42

• The expert system is in a constant state of awareness (i.e. it will not

overlook any critical events that it has been assigned, and has the means to
monitor

• It makes its decisions with regard to the goals of the firm, rather than with
regard to how such decisions might influence its personal promotions or
pay raises

• The expert system multiplies the expertise of the firm, that is, it is directly
accessible by all other divisions of the firm whereas the human expert's
access is limited by physical and geographical considerations.

• The expert system is, itself, a repository for the storage of the knowledge
of those experts from whose input it was developed; it is a knowledge
bank of considerable value and is thus a tangible, and pennanent asset of
the firm.

3.2 History of Expert systems

The technology represented by current expert systems is an outgrowth of the AI
techniques that have been the subject of intensive research since the late 1950s.

Alberico and Micco'" summarize the major milestones in the history of expert
systems developments in the following manner.

First Generation, Pre-1956

Warren McCulloch is considered to be one of the founding fathers of AI. At the
University of Illinois, he had developed a neural net model of the brain. Though
this model, later proved fundamentally flawed on various aspects, its influence
was substantial.
Chapter 3: Overview of Expert Systems

Turing was the first one to visualize the 'think machines.' He is remembered in
the Al circles for his famous 'Turing Test.' In this test, an operator interacts with
two parties, one a human and the other a computer. If the operator fails to
understand whether he is interacting with a computer or the human at any given
time, it is assumed that the computer possess human reasoning.

Second Generation, 1956-1970

John McCarthy of Dartmouth College and his friend Marvin Minsky of MIT
organized for the first time, a ten-man study of Al in 1956 at Dartmouth. This also
facilitated the term "Artificial Intelligence" to appear in print for the first time.
This conference is a milestone in the history of AI. Alan Newell and Herbert
Simon from the Carnegie Institute of Technology, around the same time, were
working on the 'Logic Theorist', which later evolved into the 'General Problem
Solver.'

During this period some early expert systems with limited capabilities were also
developed by AI researchers. BAGGER - a rule based system designed to pack
groceries in bags and ANALOGY - an expert system to perfonn the I.Q. test were
some of them. One of the major systems to be built in this period was the
MACSYMA by two MIT researchers. MACSYMA was a powerful problem-
solving system with the capability of performing over 600 distinct mathematical
operafions.

DENDRAL, one of the most blown expert systems even today was built by
Edward Feigenbaun in 1968. The system helps in identifying the structure of a
parent compound from an input fonnula.
Chapter 3: Overview of Expert Systems 44

1970 -1982
Many large systems were developed during this period. The approach was slowly
shifted from problem solving to modelling of human expertise. Some of the well
known systems like PROSPECTOR, MYCIN and XCON were developed in this
period.

1982 - onwards
This is considered to be the golden period for expert systems research. Countries
like the U.S. and Japan had veiled various new programmes with large funding.
A number of computer vendors jumped into this area and many commercial
expert system shells were made available. A large number of expert systems have
been developed in various fields performing a range of different tasks.

3.3 Architecture of an expert system

Figure 3.1 explains the basic architecture of an expert system. The user interacts
with the system through a user interface which may use menus, natural language
or any other style of interaction. Then an inference engine is used to reason with
both the expert knowledge (extracted from an expert) and data specific to the
particular problem being solved. The expert knowledge will typically be in the
form of a set of IF-THEN rules. The case specific data includes both data
provided by the user and partial conclusions (along with certainty measures)
based on this data. In a simple forward chaining rule-based system the case
specific data will be the elements in working memory.
Chapter 3: Overview of Expert Systems 45

Advice

User Interface (
User
^ "-,
ivcqucsis lor auvice
Requests for Observed data
values and advice Values of variable
1' advice
Interface
Engine
k
Requests
For Rules 1lules

Knowledge
Base

Figure 3.1 Architecture of an Expert System (adapted from Yazdani, Masoud", 1984)

The main components'^ of an expert system are:

1. the knowledge base
2. the inference engine
3. a knowledge-acquisition module
4. an explanatory interface.

Each of these is discussed in detail in the following sections.

3.3.1 Knowledge base

The knowledge base stores information about the subject domain. Knowledge is
stored in a logical manner so that the inference engine can perform deductions
upon it. The specification knowledge base contains information about particular
Chapter 3: Overview of Expert Systems 4g

applications not necessarily disjoint, especially if these applications are related to

different aspects of the same management environment. This information is
covered by clusters of connected frames and captures the presence or absence of
contradictions in the ground knowledge. Each application models information
about a system or part of it'^

There are various methods to represent knowledge in an expert system. Following

are the major knowledge representation methods'''•.

Rules in IF-THEN format

The condition specifies some pattern and the conclusion may be an action or
assertion. In such a system, the domain knowledge is represented as a set of rules
that are checked against a collection of facts or knowledge about the current
situation. When the IF portion of the rule is satisfied by the facts, the action
specified by the THEN portion is performed. When the condition is satisfied the
mle is said to 'fire' or 'execute'. A njle interpreter is used to compare the IF
portions of rules with the facts and execute the mle whose IF portion matches the
facts. The rule's action may modify the set of facts in the knowledge base.

Semantic nets
These represent relationships among objects in the domain by links between
nodes. A semantic network is a method of knowledge representation based on a
network structure. A semantic net contains points called nodes connected by links
called arcs. The nodes represent objects, concepts or events. The arcs are used to
represent the relations between the nodes. Arcs may be employed for varied
purposes. The most common arcs represent the knowledge in hierarchies.
Chapter 3: Overview of Expert Systems 47

Frames
These are generalized record structures which are more flexible than is usual in
traditional data-processing in that they may have default values or procedural
entries as the values of certain fields or slots. Frames are in a way abstractions on
groups of facts. Frames play an important role in dealings with potentially large
number of facts, because they help organize facts. Frames are something called
classes or prototype or structured object descriptions. The objective of a frame
system is to partition data, just as modular programming partitions programs. The
usual way is to group together facts about the same thing or object, or grouping
facts with the same argument values.

Horn clauses
These are expressions in a form of predicate logic on which Prolog is based and
with which the prolog system can perform inferences.

In the present work, we have used a rule-based approach for knowledge

acquisition. Rules are the most predominant fonn for knowledge representation.
MYCIN, one of the first expert systems to be built was based on this method. The
rule based systems start with a set of rules, which contains all the necessary
knowledge encoded in the IF-THEN format and a working memory. The system
will examine all the rule conditions (IF), and determines a subset, the conflict set,
of rules whose conditions are satisfied based on the working memory. Of this
conflict set, one of those rules is triggered. Wnich one is chosen is based on a
conflict resolution strategy. When the rule is fired, any actions specified in its
THEN clause are carried out.
Chapter 3: Overview of Expert Systems 48

Following is an example of iiile-based knowledge representation model'°:

IF 1. the infection is meningitis, and

2. the sub-type of meningitis is bacterial, and
3. only circumstantial evidence is available, and
4. the patient is atleast 17 years old, and
5. the patient is alcoholic

THEN
there is suggestive evidence that Diplococcus pneumoniae is an organism
causing the meningitis

As already mentioned, for the present work, we have adopted a rule-based

knowledge representation model. The knowledge base contains the AACR2 rules
in an IF-THEN format.

3.3.2 The inference engine

An inference engine tries to reach conclusions or inferences based on the data
available in the knowledge base. It is the most important part of the expert
system. Most of the inference engines use one of the two reasoning strategies -
for\vard chaining or backward chaining.

Forward chaining starts with the data available and uses the inference rules to
conclude more data until a desired goal is reached. An inference engine using
forward chaining searches the inference rules until it finds one in which the IF-
clause is known to be true. It then concludes the THEN-clause and adds this
information to its data. It would continue to do this until a goal is reached.
Because the data available determines which inference rules are used, this method
is also called 'data driven.' The famous CLIPS expert system developed by
Chapter 3: Overview of Expert Systems 49

Charles Forgy for the NASA-Johnson Space Centre used this method. Also, Jess,
the expert system for the Java platfonn developed by Ernest Friedman-Hill of
Sandia National Labs used this method.

The backward chaining starts with a list of goals and works backwards to see if
there is any data which will allow it to conclude any of these goals. An inference
engine using backward chaining would search the inference rules until it finds one
which has a THEN-clause that matches a desired goal. If the IF-clause of that
inference rule is not known to be true, then it is added to the list of goals. Prolog,
the programming language supports this method. One of the the famous expert
systems, MYCIN was built based on this method of inferencing.

3.3.3 Knowledge acquisition

Acquiring expert Icnowledge is considered to be one of the difficult tasks in
developing expert systems. The main reason is that the experts are usually very
poor in describing their reasoning process for a particular problem. Forsyth'^ says
it is difficult to get the knowledge out of the experts for any of the three reasons -
(a) deliberate resistance, (b) inarticulacy, or (c) cognitive mismatch between
brains and programs.

Many techniques have been developed to help elicit knowledge from an expert.
These are referred to as knowledge elicitation or knowledge acquisition (KA)
techniques. The tenn "ICA. techniques" is commonly used in this context.
Scientists have also tried automating the entire knowledge acquisition process
since it is very difficult to acquire knowledge from the experts .

Ontologies are used these days for the knowledge acquisition process. Although
there is a lack of unanimity in the exact definition of the term ontology, it is
generally regarded as a formalized representation of the knowledge in a domain
Chapter 3: Overview of Expert Systems 50

taken from a particular perspective or conceptualization. The main use of an

ontology is to share and communicate knowledge, both between people and
between computer systems. A number of generic ontologies have been
constructed, each having application across a number of domains which enables
the re-use of knowledge.

3.3.4 The interface

The interface is meant for communication with the user. The user interface is
generally not a part of the ES technology, and was not given much attention in the
past. However, it is now widely accepted that the user interface can make a
critical difference in the perceived utility of a system regardless of the system's
performance. The user interface checks all responses to ensure that they are of the
correct data type. Any responses that are restricted to a legal set of answers are
compared against these legal answers. Whenever the user enters an illegal answer,
the user interface informs the user that his/her answer was invalid and prompts
him/her to correct it.

Modem day expert systems incorporate concepts from human-computer

interaction research to design better interfaces.

3.4 Applications of expert systems

Expert systems are used in diverse fields like manufacturing, medical diagnosis,
software engineering, library and information services, military applications, etc.
DENDRAL, considered to be the very first expert system was used for the
identification of chemicals. It was built in the mid-1960s at the Stanford
University'^. In 1969, Camegie-Mellon University had developed HEARSAY-I
in an attempt to demonstrate the possibilit>' of speech recognition systems. An
advanced version HEARSAY-II was developed in 1971.
Chapter 3: Oven'iew of Expert Systems 51

MYCIN an expert system in blood infections, is one of the most widely known
expert system, was built 1976. The knowledge base of MYCIN contains the
heuristic rules used by physicians in the identification of certain infections'^.

General Electric Company had developed an expert system, known as

DELTA/CATS-1, in the 1980s to aid the maintenance of diesel-electric
locomotives'^.

AT&T Bell Laboratories had developed a LaSSIE, an expert system that uses a
frame-based description language and classification interference to facilitate a
programmer's discovery of the structure of a complex system^". The system is
designed to be used in a formulate-retrieve-reformulate cycle. If the answer to an
initial query is unsatisfactory, the user can reformulate the query in a variety of
ways and try again.

John V. Thomason^' at the Commonwealth Scientific and Industrial Research

Organization, Australia had developed an expert system which analyses possible
ways of water leakage through window frames. This system broke new ground in
the sense that it has heavily used graphics. The most interesting part of the
system is that the knowledge is mostly represented in the fonn of objects instead
of data.

SAL (System for Asbestos Litigation) was meant for the legal decision making. It
is a rule-based expert system. It is supposed to be most widely used expert
system in the area of law'^.

Now a days, expert system shells are being offered by commercial vendors for
various applications. One could you use these shells for developing expert
systems without having to apply much programming skills.
Chapter 3: Overview of Expert Systems 52

3.4.1 Applications of expert systems in Library and Information

(LIS) work
Similar to its applications in other fields, expert systems are developed for various
library and infonnation related activities. Expert systems have a variety of
applications in the area of Library and Information Science. Expert systems are
rapidly becoming part of the Librarian's vocabulary. They offer a level of
technology sophistication that has never been experienced in the library world.
Already work is in progress in areas such as reference work, cataloguing,
classification, indexing and intelligent information retrieval. Alberico and
Micco^'^ observe that:

In libraries there are a number of possible applications, such as the

provision of reference assistance to users as well as a variety of other
information management tasks, including acquisitions, cataloguing,
indexing, database searching, interlibrary loan, and automated
information retrieval.

According to Gibb and Sharif" , the main areas of expert system applications to
library and information science can be in the following areas:
• Classification
• Cataloguing
• Intelligent front-ends to databases
• Referral assistance
• Abstracting and summary systems
• Archives

Although expert systems can and have been applied to many areas within Library
and Information Science, the most popular areas have been online retrieval,
reference and referral serv'ices^^.
Chapter 3: Overview of Expert Systems 53

Prasad and Rath^^ have developed an expert system to capture bibliographic data
from the title pages and verso of the title pages of documents. The system uses a
heuristics methodology which can help in developing a program to automatically
identify the various bibliographic data elements. The title pages and back of the
title pages are scanned and stored as HTML pages using optical recognition
software. The heuristics are then applied on the HTML pages.

Clarke and Cronin discusses about certain early expert systems developed for
bibliographic searches. National Library of Medicine, USA has developed an
expert system for searching the MEDLINE database in 1981. A hospital in
Boston, Massachusetts has developed an expert system called PAPERCHASE for
searching the medical literature. Crystal and Jacobson^^ describe a system called
FRED developed for searching large databases. The FRED system helps the user
to select the target database based on the information provided in the user query.

The Defence Metallurgical Research Laboratory, Hyderabad, India had developed

an expert system for the classification of technical documents using the UDC
scheme^^. The system uses the UDC rules as its knowledge base. In order to suit
the automated system, the rules have taken an enumerative structure instead of
analytico-synthetic structure. The system interacts with the classifier making
him/her conform to the route suggested by the classification scheme and alerts the
classifier whenever some minor variations are found. In this way the system tries
to bring consistency in the classification process.

In 1986, the National Agricultural Library of the United States' Department of

Agriculture (USDA) developed a system called ANSWERMAN for assisting in
answering questions related to a wide variety of reference books and other
reference materials in the area of agriculture^^. The shortcomings in
ANSWERMAN were removed and modified version known as AQUAREF was
Chapter 3: Overview of Expert Systems 54

developed. This new model allowed the users to build links between knowledge
bases representing their own collections and expertise which could be chained to
other knowledge bases. It used a concept called "distributed expertise" to help the
user in finding correct answers to their queries'''.

McDonald and Weckert discuss a system called DISTREF developed for

referral services at the Charles Sturt University in Australia. The main
components of the system include a "knowledge editor" module for creating and
maintaining concept hierarchies and a "knowledge integrator" module for
building stnactural relationships among concepts.

Ohio State University has developed an expert system called SOURCEFINDER

to support the reference serxdces at the main Libraiy. Reference librarians use
SOURCEFINDER as a training aid, in unfamiliar subject areas and same is also
used when trained librarians are unavailable. The system provides direct access
to reference knowledge by means of a series of questions known as 'Question
Units' on a variety of subjects, where instructions and sources area are arranged in
the form of types of reference questions. The system has an inference mechanism
which converts reference questions into the subjects already represented in the
library catalogue and then matches the same with the resources available in the
library.

Madalli'^'' has developed VYASA, an expert system for automatic updation of

colon classification schedule. The system had used a frame based knowledge
representation model.

Wei and Cole^^ have developed a prototype web-based electronic database

selection expert system. The prototype system is designed to assist users in
selecting from among the article databases, directories, statistical and government
document electronic resources (including Web and CD-ROM bibliographic/full-
Chapter 3: Overview of Expert Systems 55

text resource databases, but excluding the online catalogs, electronic books,
encyclopaedias, and dictionaries). Users were able to select resources by any of
three methods:

(1) searching for relevant resources using free-text keywords and phrases;

(2) browsing available databases by subject categories; or

3) choosing desired database characteristics from menus (e.g. material types

indexed, chronological coverage).

MacLennan describes a system called BROWSER, designed for aiding the

librarians in literature selection in libraries. The system displays the features of
an expert, or knowledge-based system, in that it implements a set of rules, or
heuristics, which are obtainable with relative ease from a human "expert", or
which may at least be derived by a process of trial and error from a series of
postulated rules which can be refined, by the adjustment of weightings, to
reproduce the operation of such rules. It was developed using the Prolog
programming language.

Gopinath and Prasad" discuss an expert system called PROLOGOMENA,

developed to automatically generate surface structure of subject indexing strings
based on deep structure derived according to POPSl.

Panigrahi^^ discusses VISWAMITRA, an expert system for generating automatic

classification numbers based on the Colon Classification scheme. The system
uses a frame-based knowledge representation model.
Chapter 3: Overview of Expert Systems 55

3.4.2 Application of expert systems in cataloguing

Cataloguing is certainly one of the most complex activities in the library. A

cataloguer is supposed to know many rules, whatsoever the codes he/she uses,
interpret them right in various situations to make correct entries. Since
cataloguing is an intellectual exercise it requires the involvement of professional
staff. So this makes the entire process an expensive affair for any librar>'. Hence
there is a great demand for atleast partially, if not fully, automating the
cataloguing process.

Cataloguing is again one of the oldest library techniques. Catalogues were earlier
prepared only on the basis or rules of rhumb without taking into consideration the
functions to be performed by a catalogue. The present day codes such as AACR2
contain rules formulated in a systematic manner based on normative principles. A
cataloguer prepares catalogues of documents for the purpose of recording and
interpreting them for its potential users. He/she prepares the catalogue based on
the rules given in cataloguing code. Cataloguing consists of various processes
involved in the preparation of entries and maintenance of a catalogue. Each entry
is designed for satisfying a particular approach of a user.

Recent attempts to automate cataloguing through expert systems have focussed on

descriptive cataloguing because it is considered to be a rule-based activity. There
are two approaches for applying artificial intelligence techniques to cataloguing ,
basically:

(a) A human-machine interface, where the intellectual effort is divided

between the intermediary and the support system.

(b) An expert system with full cataloguing capability linked into electronic
publishing system, so that as a text is generated online, it can be passed
Chapter 3: Overview of Expert Systems 57

through a knowledge based cataloguing process without any intellectual

input from an intenTiediar>'.

The motivation of decreasing cataloguing costs by minimizing the role of the

professional cataloguer in the cataloguing has led to experiments in the
application of expert systems to cataloguing""^. The MITINET/MARC system
developed in 1986 is considered to be one of the first expert systems for
cataloguing. The system helps the user with instructions for entering
bibliographic data and translates that information into the appropriate MARC
fonnat. The system had many loopholes and it was hardly better than the
cataloguing modules in any standard integrated library systems.

MAPPER an expert system developed as part of a doctoral research was based on

a rule-based expert system concept. The design of MAPPER was quite similar to
the MYCIN expert system which was used for medical diagnosis. Using AACR2
cataloguing rules for maps and knowledge contributed by expert cataloguers, the
system was envisioned to require a cataloguer enter specific information about the
item being catalogued. The system interprets the data based on various rules in
knowledge base and produces descriptive cataloguing entry with opportunity for
the cataloguer to verify the information.

Roland Hjerppe and colleagues at the Linkoping University in Sweden had

developed an expert system called ESSCAPE (Expert Systems for Simple Choice
of Access Points for Entries). The system was developed for the cataloguing of
books and primarily concentrated on Chapter 21 of AACR2". ESSCAPE
incorporated the rules in part 2 of AACR2 formulated as a series of IF-THEN
rules and was also designed to interface with a human user through a series of
prompts'*^.
Chapter 3: Overview of Expert Systems 58

De Silva and Zainab"*'^ explain a system they have designed to educate novice
cataloguers in creating bibliographic records for published conference
proceedings as well as to improve conventional instruction in the cataloguing of
conference proceedings. The system, The Conference Proceeding Advisor System
(CoPAS), uses the relevant AACR2R rules that were identified based on the nine
types of published conference proceedings. Private knowledge or heuristics was
elicited from three human cataloguers though a multiple observation method.
This elicited personal knowledge was then modelled into a mental map of their
thought processes on how to provide a bibliographic description for published
conference proceedings. Based on the mental mappings of the experts, the expert
advisor system was developed. The development tool used was Asymetrix
ToolBook-2.

CATALYST'*'', an expert system designed to give assistance to cataloguers, was

based on a frame-based model. The system had used the popular expert system
shell, ESP-Advisor. It is a menu driven system where the burden on the user is
relatively low. The number of options from which a user has to choose is rarely
more than 6, and is often reduced to a simple yes/no response.

Davies and James"*^ developed an expert system based on frame-based model. The
system required a cataloguer to choose options from a menu about the type of
publication under consideration and access points. After a series of questions
about the publication, the system offers a template, called a frame, for the
cataloguer to record the remaining data for bibliographic description.

Ling-Hwey Jeng had made immense contributions to the field of expert systems
applications to cataloguing. In 1986, she has worked on an expert system model
to the intellectual process involved in conceptual and logical levels for
determining the title proper of monographs'*^. She further studied the general
Chapter 3: Overview of Expert Systems 59

rules of AACR2R (1998) to identify the definition and functions of a rule'*^. For
each of the rules required for cataloguing monographs, she created
condition/action pairs for rules. One of the findings of her study is that it is
difficult to implement all AACR2 rules in a single system. Hence she has
suggested building multiple subsets of rules of AACR2 which can be integrated.

Following are some of the major concerns expressed by researchers who have
tried building expert systems based on AACR2:

1. AACR2 code poses problems that seem to stem from inherent

inconsistencies and structural deficiencies'*^

2. All of AACR2 is probably too large to set up as one knowledge base in a

feasible expert system. Therefore, separate expert system modules could
be implemented for different types of documents, to be called upon as
needed'*^.

3. Many of the AACR2 rules are not explicit and many a times require the
developer to add additional rules to represent such implicit knowledge^*'

4. The inforaiation required to make cataloguhig decisions is missing for

many instances in AACR2. Rules are not particularly amenable for
immediate computerization^'.

However, the optimism of researchers like Ann Clarke and Blaise Cronin are
quite evident from their following statement^^:

Cataloguing is a rule based activity. The Anglo-American Cataloguing

Rules constitute the totality of guidance required in order to catalogue
Chapter 3: Overview of Expert Systems 50

successfully any given item cataloguing would therefore seem

suitable for an AI application.

3.5 REFERENCES

1. Winston, Patrick H. The commercial debut of Artificial Intelligence. In :

Quinlan, J. Ross (ed.). Applications of Expert Systems. Sydney : Turing
Institute Press, 1987, p.3

2. Alan Turing. Computing machinery and intelligence. In : Mind.

LIX(236), October 1950, pp. 433-460. Available at
http://turinRtest.area51 .ipupdater.com Browsed on September 2005.

3. Prasad, A.R.D. Introduction to Artificial Intelligence. In : Proceedings of

the DRTC Refresher Workshop on Artificial Intelligence and its
applications to Library and Inforaiation work. Documentation Research
and Training Centre, Bangalore, 1993. Paper A.

4. Winston, Patrick H. The commercial debut of Artificial Intelligence. In :

Applications of Expert Systems. Edited by J. Ross Quinlan. Sydney :
Turing Institute Press, 1987, pp 4-5.

5. Borko, Harold. Getting started in libraiy expert systems research. In :

Information Processing & Management. 23(2), 1987, p.81

6. Ignizio, James P. Introduction to Expert Systems: The development and

implementation of ruie-based expert systems. New York : McGraw-Hill,
1991,p.U
Chapter 3: Oven'iew of Expert Systems

7. Hayes-Roth, Frederick, et al. An overview of expert systems. In :

Building Expert Systems. Ed. By Hayes-Roth, Frederick, et al. London :
Addison-Wesley Publishing Company, 1983, p.5

8. Waterman, Donald A. A guide to expert systems. Harlow : Addison-

Wesley, 1999, p.7

9. Ignizio, James P. Introduction to Expert Systems: The development and

implementation of rule-based expert systems. New York : McGraw-Hill,
1991, p.9

10. Aiberico, Ralph and Micco Mary. Expert systems for reference and
information retrieval. London : Meckler, 1990, pp.22-30.

11. Yazdani, Masoud. Building an expert system. In : Forsyth, R. (ed.).

Expert systems: Principles and case studies, 2nd edition. London :
Chapman and Hall, 1984, pi75.

12. Forsyth, R. (ed.). Expert systems: Principles and case studies, 2nd edition.
London : Chapman and Hall, 1984, p.9.

13. Prasad, A.R.D. Applications of computer-based language processing

tools and techniques in developing indexing languages. Ph.D Thesis.
Kamataka University, 1993. (Unpublished).

14. Forsyth, R. (ed.). Expert systems: Principles and case studies, 2nd edition.
London : Chapman and Hall, 1984, p. 11.
Chapter 3: Overview of Expert Systems 62

15. ibidp.17

16. Hart, A.E. Experience in the use of an inductive system in knowledge

engineering. In: Bramer, M. (ed.). Research and development in expert
systems. Cambridge : Cambridge University Press, 1985, p.40

17. Ignizio, James P. Introduction to Expert Systems: The development and

implementation of rule-based expert systems. New York : McGraw-Hill,
1991, pp. 51-52

18. ibidp.53

19. Bonissone, P.P. and H.E. Johnson. Expert system for diesel-electric
locomotive repair. Knowledge-Based Systems Report, General Electric
Company, New York, 1983.

20. Devanbu, Premliumar, et al. A knowledge-based software information

system. In : Proceedings of the Eleventh International Joint Conference
on Artificial Intelligence, 20-25 August 1989, sponsored by the
International Joint Conference on Artificial Intelligence, Inc., Vol. 1,
pp.110-115.

21. Thomson, John V. A water penetration expert system using PROLOG

with graphics. In : Quinlan, J. Ross (ed.). Applications of Expert Systems.
Sydney : Turing Institute Press, 1987, pp.48-65.

22. Waterman, Donald A., et al. Expert systems for legal decision making.
In : Quinlan, J. Ross (ed.). Applications of Expert Systems. Sydney :
Turing Institute Press, 1987, pp. 23-47
Chapter 3: Overview of Expert Systema 53

23. Alberico, Ralph and Micco Mary. Expert systems for reference and
information retrieval. London : Meckler, 1990, p.3

24. Gibb, Forbes and Sharif, Carolyn. CATALYST: an expert assistant for
cataloguing. In : Program. 22(1), January 1988, p.62

25. Williams, James G. and Sochats, Ken. Application of expert

agents/assistants in Libraiy and Infoniiation systems.
http ://ltl 13. exp. sis .pitt. eduAVebsiteAVebresume/Expert AgentsPaper/Exper
t.htm
Browsed on August 2005

26. Prasad, A.R.D. and Rath, Durga Sankar. Heuristics for identification of
bibliographic elements from verso of title pages. In : Library Hi Tech.
22(4), 2004, pp.397-403.

27. Clarke, Ann and Cronin, Blaise. Expert systems and library/information
work. In : Journal of Librarianship. 15(4), 1983, p.286-7.

28. Crystal, M.L, and Jacobson, G.E. FRED, a front end for databases. In :
Online. September 1982, pp.27-30.

29. Gowtham, M.S. and Kamat, S.K. An expert system as a tool to

classification. In : Library Science with a Slant to Documentation and
Information Studies. 32(2), June 1995, pp.57-63.

30. Samuel T. Waters. Answerman, the Expert Information Specialist: An

Expert System for Retrieval of InfoiTnation from Library Reference
Chapter 3: Overview of Expert Systems 64

Books. In : Information Technology and Libraries. September 1986, pp.

204-212

31. Deborah Hanfman. AquaRef: An Expert Advisory System for Reference

Support. In : The Reference Librarian. No.23, 1989, pp. 113-133

32. McDonald, Craig and Weckert, John. Libraries and expert systems.
London : Taylor Graham, 1991.

33. Stalker, J.C. and Murfin, M.E. Why reference librarians won't
disappear: a study of success in identifying answering sources for
reference questions. In : RQ. 35(4), Summer 1996, pp.489-503.

34. Madalli, Devika P. A knowledge representation model for management

of analytico-synthetic scheme: a case study of Colon Classification. Ph.D
Thesis. University of Mysore, 2000. {Unpublished).

35. Ma, Wei and Cole, Timothy W. Genesis of an electronic expert system.
In : Emerald FuUtext database.
http://www.emeraldinsight.com/Insight/htmI/Output/Published/EmeraldFu
llTextArticle/Articles/2400280301.html
Browsed on April 2005

36. MacLennan, Alan. Fiction selection - an AT at work? In : Emerald

FuUtext database.
http://www.emeraldinsight.com/Insight/html/Output/Publish
ed/EmeraldFuilTextArticle/Articles/0720970204.hti-nl
Browsed on April 2005.
Chapter 3: Overview of Expert Systems 65

37. Gopinath, M,A. and Prasad, A.R.D. A knowledge representation model

for analytico-synthetic classification. In : Proceedings of the 3rd
International Society for Knowledge Organization (ISKO) Conference:
Knowledge organization and qualitj' management Copenhagen, Denmark,
20-24 Jun 94. Edited by Harme Albrechtsen and Susan Oemager,
Frankfurt/Main, INDEKS Verlag, 1994, p.320-7.

38. Panigrahi, Pijustikanti. Design and development of a knowledge

representation model for analytico-synthetic classification. Ph.D. Thesis.
Vidyasagar University, 1996. {Unpublished).

39. Clarke, Ann and Cronin, Blaise. Expert systems and library/information
work. In : Journal ofLibrarianship. 15(4), 1983, p.289.

40. Anderson, Barbara. Expert systems for cataloguing: Will they

accomplish tomorrow the cataloguing of today? In : Cataloguing &
Classification Quarterly. 11(2), 1990, p.33

41.ibidp.38

42. Hjerpe> Roland and Olander, Birgitta. Cataloguing and expert systems:
AACR2 as a Knowledge base. In : Journal of American Society for
Information Science. 40(1), 1989, pp.27-44

43. De Silva, Sharon Manel and Zainab, Awang Ngah. An advisor for
cataloguing conference proceedings: design and development of CoPAS.
In : Cataloguing and Classification Quarterly. 29(3), 2000, pp.63-80.
Chapter 3: Overview of Expert Systeni^i 66

44. Gibb, Forbes and Sharif, Carolyn. CATALYST: an expert assistant for
cataloguing. In : Program. 22(1), January 1988, pp.62-71

45. Davies, Roy and James, Brian. Towards an expert system for
cataloguing: some experiments based on AACR2. In : Program. Vol.18,
1984, pp.283-97

46. Jeng, Ling-Hwey. An expert system for determining title proper in

descriptive cataloguing: a conceptual model. In : Cataloguing and
Classification Quarterly. 7(2), 1986, pp.55-71.

47. Jeng, Ling-Hwey. The stmcture of laiowledge base for cataloguing rules.
In : Information Processing & Management. 27, 1991, pp.97-110.

48. Hjerpe, Roland and Olander, Birgitta. Cataloguing and expert systems:
AACR2 as a Knowledge base. In : Journal of American Society for
Information Science. 40(1), 1989, p.36

49. ibid p.38

50. Davies, Roy and James, Brian. Towards an expert system for
cataloguing: some experiments based on AACR2. In : Program. Vol.18,
1984, p. 288

51. Sauperl, Alenka and Saye, Jerry, D. Pebbles for the mosaic of
cataloguing expertise: what do problems in expert systems for cataloguing
reveal about cataloguing expertise? In : LRTS. 43(2), 1998, pp.90-91
Chapter 3: Overview of Expert Systems 67

52. Clarke, Ann and Cronin, Blaise. Expert systems and library/infomiation
work. In : Journal of Librarianship. 15(4), 1983, p.289.