Development of a systematic approach for knowledge acquisition and experience capture of veteran practitioners in the hi...


Material Information

Development of a systematic approach for knowledge acquisition and experience capture of veteran practitioners in the highway construction industry
Physical Description:
xiii, 316 leaves : ill. ; 29 cm.
Epstein, William C
Publication Date:


bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1995.
Includes bibliographical references (leaves 310-315).
Statement of Responsibility:
by William C. Epstein.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002070226
notis - AKQ8489
oclc - 34428423
System ID:

Full Text






Dedicated to Jim


Only the good die young, and you my brother
were the absolute very best. Say HI to mom,
and remember always, that I love you forever.

With all my heart and soul,

great lover of people and mu-
sic, passed away peacefully
at the age of 40, on Aug. 10,
1995. He has gone on to join
his mother, Edie, and is sur-
vived by wife Jill, son Justin,
father David, brothers Bill
and Bob, and sister Carolan,
and other loving family mem-
bers and friends. Graveside
Services to be held Sunday.


I would like to express my deepest gratitude to Dr. Zohar Herbsman, my mentor and

supervisory committee chairman, for all of his wisdom, guidance, and encouragement, both

on a professional level and a personal one. As for Dr. Ralph Ellis, my committee cochairman,

his dedication to the profession of academia is an inspiration. I would also like to take this

opportunity to thank my external committee member, Dr. Leon Wetherington, as well as my

other committee members, Dr. Paul Thompson and Dr. David Bloomquist, for their

continued support during my tenure at the University of Florida.

I would be remiss in not acknowledging the efforts of the many Florida Department

of Transportation (FDOT) personnel who gave freely of their time in assisting this research

endeavor. The members of the FDOT District 2 construction offices in both Lake City and

Jacksonville deserve special recognition, as their input and expertise was particularly

valuable. A special thank you also goes out to Ana Maria Elias and Daniel Baudino for all

of their assistance with respect to the IN REACH prototype system development.

Additionally, their friendship and humor helped me keep my sanity in the closing days of

writing this dissertation. I would also like to thank the makers of Cafe Bustelo; they too

were instrumental in the final drive to complete this research effort.

And finally I want to acknowledge my family and Hilda. I want to thank my dad for

just being my dad and my sister Carolan for handling everything recently. My mom and my

brother Jim, although no longer with me in body, will always be with me in spirit, and I thank

them for the time they gave me. And last but never least, I want to thank the lovely and

beautiful Hilda, my best friend and my soon to be bride.




ACKNOW LEDGM ENTS ................................................. ................................ iii

LIST OF TABLES ..... .......................... ................ viii

LIST OF FIGURES ............................................................................ ................. ix

A B ST R A C T ......................................................... ............... ....... ........................ x ii


1 IN TR O D U C T IO N .......................................... ........................................... 1

1.1 General Com m ents ............................. ..... .................................. 1
1.2 Problem Statem ent .............................. ............................................ 2
1.3 Research Objectives ............................ .... ........................... 4
1.3.1 General Comments ............................................................ 4
1.3.2 Breakdown of the Research Objectives .................................. 5
1.4 Research Methodology ......................................... 6
1.4.1 General Comments ................................. ................... 6
1.4.2 Breakdown of the Research Methodology Phases .................... 6

2 SURVEY OF CURRENT PRACTICES ....................................... ..... 11

2.1 Survey of Governmental Highway Agencies ...................................... 11
2.1.1 Introduction ................................... ............... 11
2.1.2 Breakdown of the KA & EC Questionnaire ............................ 11
2.1.3 Distribution of the KA & EC Questionnaire ............................ 13
2.1.4 Rates of Response to the KA & EC Questionnaire .............. 14
2.1.5 Section by Section Results of the KA & EC Questionnaire ....... 15
2.1.6 Selected Comments from the KA & EC Questionnaire ......... 26
2.1.7 Sum m ary .................................. ........ ..................... 27
2.2 TRB Information on Current Practices ............................ ........... 29
2.2.1 TRB Synthesis on Knowledge Based Expert Systems.............. 29
2.2.2 Survey of the TRB Construction Management Committee ........ 31
2.3 Current Practices Within the U.S. Army Corps of Engineers ............. 32
2.3.1 General Comm ents ........................................ ... ............ 32

2.3.2 Jacksonville District Corps of Engineers .............................. 33
2.3.3 U.S. Army Construction Engineering Research Laboratories .... 34
2.4 Summary of the Survey of Current Practices ....................................... 40
2.4.1 The General Category of Work for Further Concentration ........ 40
2.4.2 Knowledge Acquisition and Experience Capture Methods ........ 41
2.5 Final Comments on the Survey of Current Practices ......................... 44

3 REVIEW OF PUBLISHED LITERATURE ............................ ......... 45

3.1 Introduction ........................................................ 45
3.2 Knowledge Based Expert System ................................................ 46
3.2.1 G general Com m ents ........................................ ..................... 46
3.2.2 H historical Background ......................... ....................... .... 47
3.2.3 Generalized Overview of Knowledge Based Expert Systems .... 48
3.3 Hypertext ....................................................... 54
3.3.1 General Comm ents ........................................ ... ............ 54
3.3.2 Historical Background ........................ .......... 55
3.3.3 Generalized Overview of Hypertext ........................................ 59
3.4 Database Management Systems .................................... 65
3.4.1 G general Com m ents ........................................ ..................... 65
3.4.2 H historical Background ............................................................. 66
3.4.3 Generalized Overview of Database Management Systems ........ 68
3.4.4 A Closer Look at Relational Database Management Systems ...... 74
3.5 Sum m ary and Conclusions ........................................ ...................... 77
3.5.1 General Comments ........................................ .......... ... 77
3.5.2 Considerations Regarding Proposed Integrated Environment .... 78
3.5.3 Software Requirements for Prototype System ..................... 82
3.5.4 Final Selection of Software Package for Prototype System ....... 83
3.5.5 Final Comments ........................................ 84


4 .1 Introduction .......................................... .............................. .......... 86
4.2 The Traditional Approach to Knowledge Acquisition ............................ 87
4.2.1 General Comments ................................................. 87
4.2.2 An Overview of the Traditional Approach ............................ 87
4.3 The IN REACH Modified Approach to Knowledge Acquisition ............ 89
4.3.1 General Comments ............................................. 89
4.3.2 An Overview of the Modified Approach ................................ 89
4.4 The IN REACH Base of Documented Knowledge and Experience ........ 91
4.4.1 General Statement ......................... ........................... 91
4.4.2 A Closer Look at the Document Base of IN REACH ........... 92
4.5 Lessons Learned from Post Construction Conferences .......................... 108
4.5.1 General Comments ........... ........ .................. 108

4.5.2 FDOT Process Performance Reviews Lessons Learned ......... 108
4.5.3 UF Post Construction Conferences Lessons Learned .............. 114
4.6 The Hierarchial Structure of the Hypertext Network of IN REACH .... 118
4.6.1 General Statement ................................................................ 118
4.6.2 A Graphical Representation of the Hierarchal Structure ........... 118
4.7 Final Comments .......................................... ................. 125


5.1 General Comments ............................................................. ............... 130
5.2 A General Overview of KnowledgePro for Windows ......................... 131
5.3 Some Programming Details About Browsing and Searching ................ 132
5.3.1 General Comments ............................................................... 132
5.3.2 Developing the Browsing Capabilities of IN REACH .............. 133
5.3.3 The Windows Resource Archive Program .............................. 134
5.3.4 Developing the Searching Capabilities of IN REACH ............. 135
5.4 A Guided Tour of the IN REACH Prototype System ......................... 140
5.4.1 Introduction ..................................................................... 140
5.4.2 The IN REACH User Interface Layout and Basic Functions .... 143
5.4.3 A Closer Look at the Three "Search By" Routines ................ 148
5.5 Testing of the Prototype System ..................................................... 159
5.5.1 General Comments ....................... .. ................ 159
5.5.2 Structured Demonstrations of Preliminary Versions ............... 159
5.5.3 Distribution of Prototype System for Unsupervised Testing .... 160
5.6 Final Comments .................................... ....... ................. 162

6 CONCLUSIONS AND RECOMMENDATIONS ........................................ 164

6.1 General Comments .................................................................. 164
6.2 A Summary of the Originally Stated Research Objectives .................... 165
6.2.1 General Comment ............................................... 165
6.2.2 Summarized Objective Number One ........................................ 165
6.2.3 Summarized Objective Number Two ...................................... 165
6.2.4 Summarized Objective Number Three .................................... 166
6.3 A Review of Whether or Not the Objectives were Accomplished ......... 166
6.3.1 General Com m ent .............................................................. 166
6.3.2 Reviewed Accomplishments as per Objective Number One ...... 166
6.3.3 Reviewed Accomplishments as per Objective Number Two ..... 166
6.3.4 Reviewed Accomplishments as per Objective Number Three .... 167
6.3.5 Final C om m ent ................................................................. 168
6.4 Recommendations for Future Enhancements to the Prototype System ... 168
6.4.1 Software Enhancements ........................................ ............. 168
6.4.2 Development of a More Sophisticated Rule Set ..................... 169
6.4.3 Incorporation of More Multimedia Features ........................... 169
6.5 Final Comments ..................................... ................. 170


A KA & EC QUESTIONNAIRE ....................... ..... .................... ................ 172








I USACERL FACT SHEETS OF TWO KBES PROGRAMS ............................. 236







R E FE R E N C E S ................................................................................... ..................... 3 10

BIOGRAPHICAL SKETCH ...................... ......................... ....... .................... 316




2.1 Results of the North American Questionnaire (Section I) ................................ 16

2.2 Results of the Florida Questionnaire (Section I) ............................................ 17

2.3 Results of the North American Questionnaire (Section II) ............................ 18

2.4 Results of the Florida Questionnaire (Section II) ............................................ 19

2.5 Results of the North American Questionnaire (Section III) ........................... 21

2.6 Results of the Florida Questionnaire (Section II) .................... ............. 22

2.7 Results of the North American Questionnaire (Section IV) ............................ 24

2.8 Results of the Florida Questionnaire (Section IV) ........................................ 25

2.9 Level ofKBES Activity Among the SHAs of the United States ..................... 30




1.1 Research Development Flowchart .................... ...................... ................ 7

2.1 Comparison of Results from North America Versus Florida (Section II) ......... 19

2.2 Comparison of Results from North America Versus Florida (Section III) ......... 22

2.3 Schematic Flowchart of ARMS Operations ............................. ................. 37

2.4 Schematic Flowchart of BCOE Advisor Operations ....................................... 39

3.1 History of the Evolution of Expert Systems from Artificial Intelligence ......... 49

3.2 Architecture of a Generic Knowledge Based Expert System ......................... 51

3.3 Hypermedia as a Fusion of Hypertext and Multimedia .................................. 61

3.4 A Hypertext Network of Three Nodes Connected by Four Links ...................... 62

3.5 Correspondence Between Display Screen and the Hypertext Database ............ 64

3.6 Typical Representation of the Hierarchal Model ............................................. 71

3.7 Typical Representation of the Network M odel ............................................... 73

3.8 Typical Representation of the Relational Model ............................................. 75

3.9 Integrated Hypertext Network With Browsing and Searching Capabilities ......... 81

4.1 IN REACH Index of Sources Index Screen for BRIDGES .......................... 93

4.2 IN REACH FDOT Standard Specs Index Screen for BRIDGES ................. 94

4.3 IN REACH FDOT Supplemental Specs Index Screen for BRIDGES ............. 96

4.4 IN REACH FDOT Standard Drawings Index Screen for BRIDGES .............. 97

4.5 IN REACH FDOT Pile Splices Illustration Screen for BRIDGES .............. 98

4.6 IN REACH FDOT CPAM Manual Index Screen for BRIDGES .................... 99

4.7 IN REACH FDOT Inspection Manual Index Screen for BRIDGES ............... 101

4.8 IN REACH FDOT Tricks of theTrade Index Screen for BRIDGES ................ 102

4.9 IN REACH FDOT Armor Joint Void Topic Screen for BRIDGES ................ 103

4.10 IN REACH FDOT Inspection Checklists Index Screen for BRIDGES ......... 105

4.11 IN REACH CRSI Placing Rebar Index Screen for BRIDGES ....................... 106

4.12 IN REACH CRSI Bar Identification Illustration Screen for BRIDGES ............ 107

4.13 IN REACH FDOT PPR Lessons Learned Index Screen for BRIDGES ............ 109

4.14 IN REACH UF PCC Lessons Learned Index Screen for BRIDGES .............. 110

4.15 IN REACH Pole Vibration Lesson Learn Topic Screen for BRIDGES ............ 112

4.16 IN REACH Pole Vibration Project Data Topic Screen for BRIDGES ............. 113

4.17a IN REACH Pile Cap Lesson Learn Topic Screen (%) for BRIDGES ............. 116

4.17b IN REACH Pile Cap Lesson Learn Topic Screen (2/2) for BRIDGES ............. 117

4.18 IN REACH Pile Cap Meeting Data Topic Screen for BRIDGES ................... 119

4.19 IN REACH Pile Cap Project Data Topic Screen for BRIDGES ................... 120

4.20 Representation of IN REACH's Embedded Hierarchal Model ........................ 121

4.21 IN REACH IC-Preformed Pile Holes Topic Screen for BRIDGES .................. 123

4.22 IN REACH FDOT Inspecting Piles Index Screen for BRIDGES ..................... 124

4.23 IN REACH FDOT Inspection Checklists Index Screen for BRIDGES ............. 126

4.24 IN REACH Index of Sources Index Screen for BRIDGES ........................... 127

4.25 IN REACH Home Screen for General Category of BRIDGES ....................... 128

5.1 IN REACH "Classification" Database for the Subcategory of Bridge Deck ....... 137

5.2 IN REACH "Configuration" Database for the Subcategory of Bridge Deck ...... 138

5.3 IN REACH W welcome Screen .................................. ............. ........... ... 141

5.4 IN REACH General Categories Screen .......................................... ...... 142

5.5 IN REACH Home Screen for General Category of BRIDGES ....................... 144

5.6 IN REACH Zoomed In View of the Activated "Search By" Window .............. 146

5.7 IN REACH Example of Activation of the "Source" Pop Up Window ............. 147

5.8 IN REACH Example of "List of Topics" Search By Routine ....................... 149

5.9 IN REACH Example of "List of Topics" Scrolled Down to the Letter "I" ...... 150

5.10 IN REACH Example of Superimposed Subcategory "Search By" Options ...... 152

5.11 IN REACH Example of the "BRIDGE DECK" Subcategory Window ............ 153

5.12 IN REACH Selected Subjects for the "BRIDGE DECK" Example ................ 154

5.13 IN REACH "Result(s) of Search" for the "BRIDGE DECK" Example ......... 155

5.14 IN REACH Return of the "Result(s) of Search" to the Standard Interface ...... 157

5.15 IN REACH "Related Topics" Example for "415-5.13 Chairs & Bolsters" ........ 158

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



William C. Epstein

December, 1995

Chairman: Dr. Zohar J. Herbsman
Cochairman: Dr. Ralph D. Ellis
Major Department: Civil Engineering

Every company in every industry faces the prospect of the loss of knowledge and

experience through the departure of key personnel. This predicament created by the loss of

veteran employees is especially acute in the highway construction industry, where frequently

the experience is either undocumented or poorly documented, and the knowledge possessed

by these people is retained exclusively as personal property. This dissertation not only

explores the difficulties associated with pursuing an approach to acquire heretofore

undocumented construction knowledge and expertise, but it also recognizes the vast amounts

of highway construction data and information that are currently available within the

transportation industry. Any concerted effort attempting to capture the construction

knowledge and expertise of a large organization, such as a department of transportation,

would be severely remiss in not taking advantage of this existing base of documented


This research endeavor represents a comprehensive study of the problems associated

with the development of a systematic approach for capturing the knowledge and experience

of a large organization, and establishing a computer delivery system for dissemination of this

encoded information. Fundamental to this delivery system is the creation of a user-friendly

computing environment that will provide an intuitive tool capable of assisting both veteran

and novice practitioners in fashioning more informed decisions concerning problems that may

arise during normal and abnormal highway construction operations.

One of the major accomplishments of this research effort, was the development of an

information management prototype system which was given the name IN REACH

(Intelligent iNformation Retrieval and Expert Advice for the Construction of Highways). IN

REACH is comprised of an underlying, fully functionally hypertext network which is

augmented by the integration of some innovative database management and expert systems

strategies. In an effort to add structure to the inherently unstructured world of a pure

hypertext system, IN REACH utilizes these integrated strategies to enhance the user's

capability of direct queries to the overall network, both statically and dynamically.


1.1 General Comments

Like many other general concepts, knowledge is a difficult term to quantify. A good

discussion of what constitutes knowledge should begin with characterization of the difference

between data and information. Commonly speaking, raw data are nothing more than a

collection of facts and figures, that by themselves lack any real significance. Only when

meaning is assigned to these facts and figures do these data evolve into information.

Knowledge, on the other hand, can be thought of as the cognitive storage of information

which is readily available for retrieval by the conscious human mind. Feigenbaum [1984]

makes a very interesting point about the relationship between knowledge and information.

He suggests that first it should be clarified that knowledge is not synonymous with

information, rather knowledge is information that has been implemented, categorized,

applied. According to Hayes-Roth, Waterman and Lenant [1983], knowledge consists of

(1) symbolic descriptions of definitions, (2) symbolic descriptions of relationships, and (3)

procedures to manipulate both type of descriptions.

Knowledge of a certain subject, in and of itself, does not constitute expertise in that

field. Expertise is a function of the skillful application of knowledge, and this skill is a direct

result of having experience in that particular domain. What differentiates a novice from an


expert is not the quantity of knowledge possessed, but rather the amount of experience using

that knowledge. More than ever, modem industry depends on the expertise of its work force

for success. No longer in today's complex world, can one man or woman possibly know

everything there is to know.

For an organization to prosper in this environment, not only must its members possess

a certain level of expertise on an individual basis, but this personal expertise must be

exchanged and transferred throughout the entire structure of the group. The wealth of

knowledge and experience accumulated by veteran employees through their years of service

is something which clearly should be utilized and taken advantage of for current operations.

Furthermore, the fact that these veteran personnel will not remain with the organization into

perpetuity suggests that, as is the case with any limited and valued possession, their

knowledge and experience must be captured and stored for future use.

1.2 Problem Statement

The research effort presented herein is focused upon the United States highway

construction industry from the perspective of the governmental state highway agencies

(SHAs). Every state in this country has a representative SHA which is responsible for the

construction of the transportation systems within their boundaries. Not unlike any other

large organization, SHAs continually face the unfavorable prospect of losing significant

amounts of accrued knowledge and experience as a result of ongoing departures of key

personnel. These veteran employees, many of whom have spent their entire professional

careers under the employ of a single SHA, aggregately represent thousands of years of


accumulated expertise. This predicament is especially acute in the realm of highway

construction operations, wherein frequently the knowledge and experience possessed by

these people is either undocumented or poorly documented, and is usually retained

exclusively as their personal property. What this implies is that, upon their departure, these

seasoned practioners will take with them the years of training and experience provided to

them by the SHA, and in return leave behind little if any of their knowledge and expertise.

To further exasperate this situation, currently and over the next several years, the

SHAs of this country are and will continue to experience an exceedingly concentrated loss

of veteran personnel. The reason why this is happening is due to the fact that many key

members of today's transportation workforce began their SHA careers during the highway

construction boom of the late 1960s and early 1970s, and unfortunately they are all

approaching retirement age at approximately the same time. This inevitable occurrence is

going to create a critical shortage of experienced practitioners. Time is therefore of the

essence for implementation of some sort of capture program that will not allow a whole

generation of highway construction experience to disappear. Failure to capture this expertise

and integrate it into the organizational and operational structures of the various SHAs will

result in an enormous loss of knowledge that may never fully be replaced. Being that

experience is such a valuable asset in the field of highway construction, research into a

methodology for securing this resource for future use is certainly a very practical and

worthwhile endeavor.

In conjunction with the development of a functional means of acquiring heretofore

undocumented construction expertise, recognition of the vast amounts of highway

construction data and information currently available within the transportation industry is


fundamental. Over the years, SHAs across the country have produced a wealth of quality

programs and publications presenting a variety of construction related topics. Any concerted

effort attempting to capture the highway construction knowledge and experience of this

country's SHAs would be severely remiss in not taking advantage of this existing base of

documented information.

Along with preserving potentially irreplaceable construction expertise and utilizing

existing data and documentation, the other critical aspect of a successful experience capture

program is the proposed method of disseminating the acquired information. This information

and the knowledge associated with this information must first be formalized and encoded

into some sort of communicable form. Then, through a computerized storage, management

and retrieval system, novice and veteran personnel alike would be able to easily refer to all

available information about a particular subject. This easy access to a wide variety of related

topics would provide the user a powerful tool from which to gather the appropriate

knowledge necessary for a more informed decision making process.

1.3 Research Objectives

1.3.1 General Comments

The overall goal of this research project was to develop a systematic approach for

gathering highway construction knowledge and experience, organizing this information,

storing it, and presenting it in such a fashion as to be readily accessible and useful to anyone

wishing to benefit from this knowledge base. This broad effort can be broken down into the

following research objectives as presented next.

1.3.2 Breakdown of the Research Objectives Objective 1

The initial objective of this research was to identify and prioritize the general

categories of highway construction work and operations wherein the loss of experience was

felt to be most critical. The area distinguished as most acute was then focused upon for

further concentrated research. Objective 2

Having determined the area of focus, the next objective was to identify and analyze

existing programs and published materials related to this selected field of concentration. The

appropriate information was then categorized and stored for incorporation into the final

system. Objective 3

A fundamental objective of this research was the development of a systematic

approach for capturing and documenting the individual knowledge and experience of veteran

personnel. This information was then combined with the data as collected from Objective 2

to create an integrated base of individual and organizational highway construction

knowledge. Objective 4

Having a functional knowledge base from which to draw from, the final objective was

the generation and subsequent prototype testing of the computerized delivery system. Key

to the creation of a useful and flexible information management system, was the development


of a highly intuitive and user-friendly environment that allows for relatively easy future

expansion to the basic system architecture.

1.4 Research Methodology

1.4.1 General Comments

At this point it should be noted that this dissertation is based on funded research

conducted for the Florida Department of Transportation (FDOT). As such, the data and

information collected and analyzed typically relate to FDOT highway construction

operations. Although this research effort concentrated on FDOT personnel and

documentation, any organization could utilize this basic approach in its generic form simply

by focusing the knowledge base development process on the needs specific to that

organization. Work on this particular study consisted of accomplishing the following phased

tasks, and is further illustrated by the schematic flowchart presented in Figure 1.1.

1.4.2 Breakdown of the Research Methodology Phases Phase 1

Several preliminary interviews with various members of the FDOT were conducted

as a means of developing a comprehensive questionnaire that would fully address the issue

of knowledge acquisition and experience capture within the highway construction industry.

The resulting survey was then distributed to all SHAs in the United States, with the exception

of Florida, as well as to each of the Canadian provincial highway agencies. In Florida, rather

then mail the survey directly to the central state office in Tallahassee, it was sent out

individually to each district office. Conclusions drawn from all returned questionnaires were

Crn <
0 w<

o a
oC_ 1 wd

a (D -O iiiinnnniii^ lB

00 m

-W Lgs 2 aLsf

i^"^s. ^r sas g s

mw 0
C13 0 z a4
04 w4 j 0-4

0e E I > I O04
"~ ; \ 1 "3 L4 *^0 K^ 'SO UJO
r ^f \ 2^ Q

^& 3 : 8 -

w oz
> 0
Z U u

Z m
C"Fi; 0

Z w


then used to identify the most acute area of highway construction operations for further

concentrated research. Phase 2

An extensive literature search was performed through a variety of state-of-the-art

electronic databases, in an attempt to uncover the most up-to-date literature written on this

subject. In addition to reviewing current publications, portions of the questionnaire from

Phase 1 were utilized to ascertain the level of similar endeavors that may be underway within

the different American and Canadian highway agencies. Furthermore, additional efforts were

made to communicate with other governmental and private organizations to identify the

possible existence of any type of knowledge acquisition and experience capture programs that

these contacted organizations may be implementing. Phase 3

Once the area of focus was selected, as described in Phase 1, a detailed review of all

related FDOT documentation was effected. Numerous FDOT publications were

accumulated, and selected information from these documents was then electronically stored

as the foundation on which the final computer delivery system would be built. Phase 4

The next phase was the development and implementation of a systematic approach

for capturing the undocumented experience possessed by veteran construction personnel.

Initially, as is standard with most knowledge acquisition programs, interviews with various

domain experts in the identified area of concentration were conducted. It was soon


discovered that although informative and useful in shaping the direction of the study, these

sessions were relatively inefficient for the task at hand, and the comments obtained were

often vague and unfocused Further into the research, a format for Post Construction

Conferences (PCC) was developed, wherein comments made by the field personnel specific

to a particular job could be integrated into the preliminary knowledge base that was being

compiled in Phase 3. Due to the fact that these meetings were centered around construction

related topics specific to the job that these people were currently working on, their

recollections and comments tended to be more thorough and useful. Additionally, the goal

of a systematic knowledge acquisition technique was more closely realized because the

proposed PCC approach minimized personality influences that are inherent in typical one-on-

one interview sessions. Phase 5

Critical to the success of this project was the identification of the requirements of the

end user. Throughout the research process, close contacts with many key FDOT

construction personnel were maintained in order to establish the specific departmental needs

that the final prototype computer delivery system must address. Conclusions drawn from the

survey of current practices and the literature review from Phase 2, coupled with a functional

understanding of the needs of the FDOT, led to the determination that a hybrid

programming environment would be the best platform for the accomplishment of developing

a flexible, user-friendly information management and retrieval system. The established

independent software technologies of expert systems, hypertext, and database management

systems were utilized in creating an integrated computer program entitled IN REACH, which


is an acronym for "Intelligent iNformation Retrieval and Expert Advice in the Construction

of Highways." Phase 6

Structured demonstration sessions for the presentation of preparatory versions of the

IN REACH program were organized for testing and validation purposes. Additionally,

executable files containing preliminary editions of IN REACH were distributed to selected

FDOT personnel for their unsupervised use. Comments and suggestions collected from these

different testing methods were evaluated and incorporated into the final IN REACH

prototype system. Phase 7

Results from the total research effort encompassing Phase 1 through Phase 6 were

analyzed, and a final dissertation presenting these findings was prepared.


2.1 Survey of Governmental Highway Agencies

2.1.1 Introduction

As previously noted in Chapter 1, the initial objective of this research was to identify

and prioritize the general categories of construction work and operations wherein the loss

of experience was felt to be most critical from the perspective of governmental highway

agencies. To this end, several preliminary interviews were conducted with a number of

FDOT Construction and Resident Engineers. These sessions were instrumental in gaining

a better understanding of the problem as seen from the viewpoints of typical SHA

construction personnel. From these interviews, a comprehensive "Knowledge Acquisition

and Experience Capture" (KA & EC) questionnaire was developed for distribution to the

various state and provincial highway agencies throughout the United States and Canada. A

copy of this survey along with a generic cover letter are included in this dissertation as

Appendix A.

2.1.2 Breakdown of the KA & EC Questionnaire Section I--Loss of Veteran Employees

The function of this section of the survey was twofold. One purpose was to

determine the level of importance that the different agencies placed on the loss of expert



knowledge due to the departure of veteran employees from the organization. The other

purpose served by this section was the quantification of the magnitude of loss with respect

to each agency. Section II--General Categories of Construction Work

Section II was designed as a way to numerically measure the effect that the loss of

experience has on different general categories of highway construction work. From the

developmental survey interviews conducted with the FDOT, the following five major

categories of construction work were identified for inclusion into the survey:

1) Bridge Work

2) Roadway Work (other than asphalt)

3) Asphalt Work

4) Signaling and Lighting

5) Maintenance of Traffic

The categories of Bridge Work, Roadway Work (other than asphalt), and Asphalt Work were

further subdivided into two separate categories, namely new construction work, and

maintenance and repair work. Additionally, the respondents were encouraged to list and rate

any other general categories of construction work that they deemed appropriate. Section III--General Areas of Construction Operations and Administration

Again from the preliminary interviews conducted with the FDOT, five major areas

of highway construction operative and administrative duties were distinguished as those that

typical construction personnel are most regularly involved in. These areas are as follows:

1) Constructability Analysis

2) Inspection

3) Quality Control

4) Construction Documentation

5) Departmental Documentation

As was the case with Section II, the respondents were asked to specify and rate any other

areas of construction operations and administration that were not listed but may apply to their

agency. Section IV--Knowledge Acquisition and Experience Capture Methodology

The final section of the questionnaire was devoted to determining the current

existence of, or future development of, techniques by which the various polled highway

agencies were attempting to acquire and capture the knowledge and experience of their

veteran practitioners. If applicable, upon completion of the collection and capture phases of

the particular knowledge acquisition methods listed, the questionnaire also requested that the

responding agency please specify what type of system or systems they utilized for storing

and distributing this collected information to the appropriate personnel.

2.1.3 Distribution of the KA & EC Questionnaire

Given that the intended focus of the research was to be on FDOT highway

construction operations, it was desired to distribute the KA & EC questionnaire in such a

fashion as to obtain a set of comparative results between North America in general and the

state of Florida in particular. With this in mind, two separate survey packages were sent out.

The North American survey package was mailed to the attention of each SHA construction


department in the United States, with the exception of Florida. Additionally, every

provincial highway agency in Canada was included in this mailing. In order to contrast these

results with those in Florida, this same survey was transmitted to each of the district

construction offices within the FDOT. Distribution lists for the North American survey

package, as well as the Florida package, are given in Appendix B.

2.1.4 Rates of Response to the KA &EC Questionnaire

The North American survey was distributed to a total of 61 agencies (the 49 SHAs

of the United States, excluding Florida; the District of Columbia Department of Public

Works; and the 11 Canadian provincial highway agencies). From this mailing, a total of 34

responses were received. Neglecting multiple respondents from a single agency, the North

American survey realized an overall response of 28 out of 61 agencies for an approximate

rate of 46%. Although this was a relatively low rate of response, it was sufficient to serve

the intended purpose of identifying and framing the problem, thus shaping the direction of

subsequent research.

In Florida, the questionnaire was distributed to each of the seven FDOT district

offices, as well as to the construction office of the Florida Turnpike Authority. It should

be noted that copies of the questionnaire were specifically mailed to three different

individuals in District 2. This occurred because District 2 served as the main personnel

resource center for the research, and as such, there were several people who expressed an

interest in participating in the study. From this mailing, 10 responses were received. Using

the same elimination approach of multiple responses from a single district, the overall

response rate from Florida's district offices was 4 out of 8, which translates to a rate of 50%.


Although the rate of response was again somewhat disappointing, the number of

questionnaires received (10) was deemed suitable for deriving comparative results.

2.1.5 Section by Section Results of the KA & EC Questionnaire Section I--Loss of Veteran Employees

Results for Section I from the North American KA & EC questionnaire are

summarized in Table 2.1, and those from the Florida survey are presented in Table 2.2. The

two numbers to focus on from this section's results are 1) the average value for the "General

Rating" of the importance of the loss of experience and 2) the average value of the "Average

Years / Person."

The "General Rating," measured on a scale of 1 to 5, with 1 being the lowest level

of importance and 5 being the highest, is an indication of the degree of significance the

respondent feels that his or her organization places on the loss of knowledge and experience

caused by the departure of veteran employees. The average value of this "General Rating"

from the North American survey was 3.39, which was slightly higher than the Florida average

of 3.20.

The second value that should be highlighted is the "Average Years / Person." This

number is important as verification that respondents were approaching the questionnaire from

the perspective of veteran personnel only. In other words, the research effort was concerned

exclusively with experienced based issues relating to those employees who possessed

significant years of service in the highway construction industry, and as such, the survey's

intent was to purposely neglect those departing personnel who had not yet accumulated

substantial years of industry experience. Referring again to Table 2.1 and Table 2.2, it is

Table 2.1 Results of the North American Questionnaire (Section I)


Responding General Veteran Employees Years of Experience
District Office Rating Lost per Year Lost per Year

Alaska (USA) 3 10 250
Alabama (USA) 2 60 2,100
Arkansas (USA) 3 ------
California (USA) 4 1,000 30,000
Colorado (USA) 4 85 2,550
Connecticut (USA) 20 240
Georgia (USA) 3 195 5,850
Idaho (USA) 4 45 900
Illinois (USA) 5 90 1,765
Kansas (USA) 5 --- --
Kentucky (USA) 3 15 500
Louisiana (USA) 1 25 625
Maryland -1 (USA) 3 14 294
Maryland 2 (USA) 4-- -
Maryland 3 (USA) 3 10 325
Maryland 4 (USA) 4 2 60
Maryland 5 (USA) 5 2 50
Maryland 6 (USA) 3 -------- ---
Mississippi (USA) 4 26 650
Missouri (USA) 3 27 960
New York (USA) 4 780 15,600
North Dakota (USA) 4 11 372
Ohio (USA) 2 30 675
Oklahoma (USA) 4 10 300
Pennsylvania (USA) 4 500 10,000
South Dakota (USA) 3 20 400
Tennessee (USA) 4 100 4,000
Texas (USA) 3 500 12,000
Virginia (USA) 3 67 1,807
Wyoming (USA) 5 10 300
Alberta (CAN) 3 143 2,750
British Columbia 1 (CAN) 3 30 750
British Columbia 2 (CAN) 2 8 240
Nova Scotia (CAN) 2 11 300
Number of Responses 33 30 30

Average Values 3.29 Average Years / Person = 25.12

Table 2.2 Results of the Florida Questionnaire (Section I)


Responding General Veteran Employees Years of Experience
District Office Rating Lost per Year Lost per Year

Florida District 2 3 15 450
Florida District 2 1 1 35
Florida District 2 4 2 65
Florida District 2 5 --- ----
Florida- District 2 5 10 330
Florida- District 3 3 3 100
Florida District 7 3 2 50
Florida District 7 3 ----- --
Florida District 7 3 5 120
Florida District 7 2 10 300

Number of Responses 10 8 8

Average Values 3.20 Average Years / Person = 30.21

apparent from the tabulated results for the "Average Years / Person" values, that both the

North American respondents who reported an average of 25.12 years / person, as well as the

Florida respondents at an average of 30.21 years / person understood the requirements of

the survey and responded accordingly. Section II--General Categories of Construction Work

The values presented in Table 2.3 (North America) and Table 2.4 (Florida) represent

the respondents' ratings of the effects of the loss of experience with respect to various

general categories of construction work as specified in Section II of the KA & EC

questionnaire. Ratings for this section were based on a scale of 1 to 10, with 1 being the

lowest level of significance and 10 being the highest. Additionally, Figure 2.1 has been

generated as a means of graphically illustrating the average values of the North American

survey in comparison with those of the Florida survey.

Table 2.3 Results of the North American Questionnaire (Section II)


Responding Bridge Bridge Road Road Asphalt Asphalt Sig & Maint of Other
Agency New R&M New R&M New R&M Light Traffic (#Only)

Alaska (USA) ---- ----- --- --- --- ----- ---- ----- -----
Alabama (USA) 6 5 6 5 7 6 7 6 ---
Arkansas (USA) 5 6 4 4 5 3 4 4 1
California (USA) 8 8 7 7 6 6 9 5 2
Colorado (USA) 5 5 5 5 5 5 5 5 -----
Connecticut (USA) 9 9 9 9 9 9 7 9 2
Georgia (USA) 5 6 5 6 7 7 3 4 ---
Idaho (USA) 8 5 9 3 8 4 7 7 1
Illinois (USA) 8 7 7 6 8 6 9 9 ---
Kansas (USA) 10 10 10 10 10 10 7 5 1
Kentucky (USA) 7 8 10 8 7 9 5 5 2
Louisiana (USA) 7 6 5 5 5 5 3 2 -----
Maryland- 1(USA) 10 6 9 5 8 4 4 7 2
Maryland-2 (USA) --- 4 --- 4 ----- 4 ----- 4 --
Maryland 3 (USA) 5 4 5 4 6 5 5 1 ----
Maryland 4 (USA) --- ----- ---- --- ---- -- --- ----- ---
Maryland- 5 (USA) ----- --- ---- --- ----- ---- 1 2 3
Maryland -6 (USA) ----- ----- ---- ----- ----- ---- ----- ---- 3
Mississippi (USA) 8 8 7 8 8 7 8 9 ---
Missouri (USA) 5 5 5 3 5 3 3 3 ----
New York (USA) 8 7 8 7 7 6 6 6 ---
North Dakota (USA) 7 ---- 7 --- 7 ---- --- 7 --
Ohio (USA) 7 6 7 5 5 4 8 5 2
Oklahoma (USA) 9 ---- 9 ----- 10 ----- -- ---- --
Pennsylvania (USA) 10 10 9 9 9 9 7 5 ----
South Dakota (USA) ----- ---- ---- ----- ----- ---- --- ---- --
Tennessee (USA) 8 5 8 5 8 5 4 4 --
Texas (USA) 9 8 8 7 8 7 8 8 3
Virginia(USA) 8 8 8 8 8 8 1 5 --
Wyoming (USA) 8 8 10 10 10 10 7 7 ---
Alberta (CAN) 8 10 7 8 6 9 4 3 1
British Columbia- 1 (CAN) 7 6 8 5 7 7 4 8 1
British Columbia 2 (CAN) 9 1 10 1 5 1 2 2 ---
Nova Scotia (CAN) 6 8 6 9 6 8 7 8 ----

Number of Responses 28 27 28 27 28 27 27 29 13

Average Values 7.50 6.63 7.43 6.15 7.14 6.19 5.37 5.34 N/ A

Table 2.4 Results of the Florida Questionnaire (Section II)

General Categories of Construction Work

Li idg I .Jrc RoJad KIJd it pli .l ,liall \ I I lkII
N.-. 1 I:& I N .-... R & I NL N '. I & I .1 L l l

North American Results Florida Results


SFigure 2.1

- Comparison of Results from North America Versus Florida ISection II


Responding Bridge Bridge Road Road Asphalt Asphalt Sig & Maint of Other
Agency New R & M New R & M New R & M Light Traffic (# Only)

Florida District 2 7 3 ----- ------ ---- ----- ----- -----
Florida District 2 7 3 ...
Florida- District 2 10 9 10 8 10 8 8 10 2
Florida District 2 9 3 9 2 8 2 6 8 2
Florida District 2 9 ----- ----- ----- 9 ----- 8 9 1
Florida District 2 8 5 8 5 8 7 5 5 1
Florida District 3 8 8 7 6 8 5 6 6 ---
Florida District 7 10 ----- 7 -8 ---- 8 7 4
Florida- District 7 8 ----- ----- 8 ---- 5 5 -----
Florida- District 7 9 10 8 1 7 2 5 6 -----
Florida- District 7 8 8 8 8 8 8 5 7 ---

Number of Responses 10 7 8 6 9 6 9 9 5

Average Values 8.60 6.57 8.13 5.00 8.22 5.33 6.22 7.00 N / A


From Table 2.3, Table 2.4, and Figure 2.1 it can be seen that the loss of experience

with respect to new construction work consistently rated as more significant than that of

repair and maintenance within the same general category of work. In the North American

survey, "Signaling & Lighting" and "Maintenance of Traffic" clearly were perceived as the

two specified categories least affected by the loss of experience. Although these two

categories on average rated higher in the Florida survey, the level of importance of these

categories with respect to the loss of veteran expertise still fell far below those categories

associated with new construction work. The column labeled "Other (# Only)" designates

other categories of work not specified in the distributed questionnaire. The numbers that

appear in this column are not ratings, rather they refer only to how many additional

categories were noted by that particular respondent. In the North American survey, the most

common "Other" category selected was landscaping. Out of 23 responses on a total of 13

different questionnaires, seven people mentioned landscaping, with an average rating of 6.17.

The "Other" category in the Florida survey, on the other hand, showed little consensus of

opinion. Section III--General Areas of Construction Operations and Administration

Section III responses are based on the same 1 to 10 rating scale as those from

Section II. Results of the North American survey and the Florida survey are summarized in

Table 2.5 and Table 2.6, respectively. A histogram that includes both sets of data has again

been included as Figure 2.2.

Analysis of the information presented in Table 2.5, Table 2.6, and Figure 2.2

demonstrated that, on average, "Constructability Analysis," "Inspection Operations," and

Table 2.5 Results of the North American Questionnaire (Section III)


Responding Constructability Inspection Quality Construct Department Other
Agency Analysis Operations Control Docs Docs (# Only)

Alaska (USA) --- 8 8 6 5 --
Alabama (USA) 7 8 8 8 7 -
Arkansas (USA) 6 5 5 4 3 1
California (USA) --- 8 9 7 9 ---
Colorado(USA) 8 3 3 5 5 --
Connecticut (USA) 8 10 9 9 9 ---
Georgia (USA) 8 5 6 4 5 ---
Idaho (USA) 9 8 7 7 7 3
Illinois (USA) 9 8 8 10 10 --
Kansas (USA) 10 10 10 5 5 1
Kentucky (USA) 10 6 8 5 5 ---
Louisiana (USA) 6 7 6 6 6 --
Maryland -1 (USA) 10 8 8 7 4 ---
Maryland 2 (USA) ----- ---- -- -
Maryland 3 (USA) 10 10 10 5 5 --
Maryland 4 (USA) --- -- -- --
Maryland- 5 (USA) 10 8 9 8 8 3
Maryland 6 (USA) 5 7 7 7 7 ---
Mississippi (USA) 9 8 8 6 7 -----
Missouri (USA) 1 5 5 5 --- --
New York (USA) 8 8 8 8 8
North Dakota (USA) 7 8 7 5 10 1
Ohio (USA) 2 8 7 6 6 -
Oklahoma (USA) 8 -- 8 ---- -
Pennsylvania (USA) 8 10 10 7 7 2
South Dakota (USA) 8 6 6 2 -- --
Tennessee (USA) 4 9 9 8 7 1
Texas (USA) 8 8 8 8 8 ---
Virginia (USA) 10 8 10 8 7 -----
Wyoming (USA) ----- 10 10 10 10 ---
Alberta (CAN) 2 10 8 5 5 ---
British Columbia 1 (CAN) 10 9 7 8 7 1
British Columbia 2 (CAN) 7 5 4 3 2 --
Nova Scotia (CAN) --- 7 7 5 5 ---

Number of Responses 28 31 31 32 29 8

Average Values 7.43 7.68 7.58 6.41 6.52 N/A

Table 2.6 Results of the Florida Questionnaire (Section III)


Responding Constructability Inspection Quality Construct Department Other
Agency Analysis Operations Control Does Does (# Only)

Florida District 2 7 7 7 5 5 -----
Florida District 2 9 10 9 8 8 1
Florida District 2 9 9 9 8 8 1
Florida District 2 ----- 8 7 8 7 -----
Florida District 2 6 7 7 7 6 -----
Florida District 3 8 9 8 8 7 2
Florida- District 7 8 7 8 9 9 I
Florida District 7 8 8 8 8 8 -----
Florida District 7 6 7 5 4 4 ----
Florida District 7 5 5 5 5 3 -----
Number of Responses 9 10 10 10 10 4

Average Values 7.33 7.70 7.30 7.00 6.50 N/ A

General Areas of Construction Operations

8 --v -- ~ -



( 0 i i-ln bIlil l 1irp'i.inll I.111jll I.. oIir.lrlnlt Ion I )ir.' nlri ll I
Analysis Operations Control Documents Documents
SNorth American Results ] Florida Results

Figure 2.2 Comparison of Results from North America Versus Florida (Section III)


"Quality Control" were thought to be those general areas of construction operations and

administration wherein the loss of experience was deemed to be most critical. With respect

to selection of "Other" areas of operations, neither the North American nor the Florida

survey yielded any definitive results. Section IV--Knowledge Acquisition and Experience Capture Methodology

Affirmative responses to the existence of the various specified knowledge acquisition

and experience capture programs are indicated by an "X" in Table 2.7, for the North

American survey, and Table 2.8 for the Florida survey. Referring to Table 2.7, the most

popular methods of acquiring knowledge, based on the North American survey, were the

"Mentor / Apprentice" approach and the use of retired veteran personnel as "Part-Time

Consultants." Results based on the Florida survey (Table 2.8) indicated that, at least among

those districts that responded, the only method that appears to be consistently utilized in

Florida is the "Mentor / Apprentice" system. Regarding applications of "Other Methods" for

acquiring knowledge, one particular technique that was mentioned by a total of five

respondents in the North American survey was the organization of training sessions

conducted by veteran practitioners. In Florida, on the other hand, no respondent gave any

information on any techniques, other than those specifically called out in the questionnaire.

With respect to dissemination of the captured construction knowledge and experience

throughout the structure of the organization, the overwhelming method of choice in both

surveys among those who chose to comment was the utilization of written construction and

inspection manuals. The Florida respondents, specifically those from District 2, commented

Table 2.7 Results of the North American Questionnaire (Section IV)


Responding on Round Depart Mentor / Post Construct Part-Time Other
Agency Interview Table Report Apprentice Conference Consultant Methods

Alaska (USA) X X X
Alabama (USA)
Arkansas (USA)
California (USA) X X X X X
Colorado (USA) X X X
Connecticut (USA) X X
Georgia (USA)
Idaho (USA) X X
Illinois (USA) X X X
Kansas (USA) X X X X
Kentucky (USA) X

Maryland -1 (USA)
Maryland- 2 (USA) X X X
Maryland- 3 (USA)
Maryland -4 (USA) X
Maryland- 5 (USA)
Maryland 6 (USA) X X X X X X X
Mississippi (USA) X X
Missouri (USA)
New York (USA) X
North Dakota (USA) X
Ohio (USA) X X X X X
Oklahoma (USA)
Pennsylvania (USA)
South Dakota (USA)
Tennessee (USA) X X
Texas (USA)
Virginia (USA) X X
Wyoming (USA) X
Alberta (CAN) X X X
British Columbia 1 (CAN)
British Columbia 2 (CAN) X X X X
Nova Scotia (CAN)

Number of Responses 6 1 2 13 9 15 9

Table 2.8 Results of the Florida Questionnaire (Section IV)


Responding 1 on 1 Round Depart Mentor / Post Construct Part-Time Other
Agency Interview Table Report Apprentice Conference Consultant Methods

Florida District 2 X X
Florida District 2
Florida District 2 X
Florida District 2 X
Florida District 2
Florida- District 3 X X
Florida District 7 X
Florida District 7
Florida District 7
Florida District 7
Number of Responses 0 1 0 5 1 0 0

on the existence of two particular in-house documents, one is a manual entitled Tricks of the

Trade [Jacksonville, 1992], and the other is a collection of inspection checklists which are

still presently under development. These documents along with several other FDOT

publications will be discussed in detail in Chapter 4 of this dissertation. In the North

American survey, there were a total of eight respondents who indicated that their

organization had developed some sort of procedural manual for highway construction

operations. An example of one such publication received through the questionnaire process

is a pocket-sized State of California Department of Transportation manual entitled Highway

Construction Checklist [State of California, 1985]. Appendix C includes selected excerpts

from this booklet, specifically, the cover, the Foreword, the Table of Contents, and the

complete section on Concrete Structures. Although this document is somewhat dated, it

does represent a comprehensive attempt by this agency at capturing the highway construction

knowledge and experience of its veteran practitioners.

2.1.6 Selected Comments from the KA & EC Questionnaire

Although the survey was rather complex and time consuming, many who participated

did take the time to give their final comments on the subject of knowledge acquisition and

experience capture as it related to their organization. The loss of valuable expertise through

the departure of veteran personnel clearly was of concern to a vast majority of the

respondents. Two comments, in particular, have been reproduced herein as an illustration

of how significant the problem is and how many perceive their agencies as highly deficient

with respect to the implementation of any type of methodology for the capture of the

construction knowledge and expertise possessed by veteran practitioners within their


The Chief of the Construction Division for the Louisiana Department of

Transportation and Development commented as follows:

I'm retiring with 33+ years of experience. Most of this has
been in structures, including all kinds of bridges and
foundation experience. When I leave, there will not be one
person in the Department that can approach my experience.
This state has done nothing, has no plans to do anything, and
probably never will address this matter.

A Resident Engineer in the FDOT made the following statement regarding the level

of importance he felt was given to this subject by his department:

The FDOT does not use any of these knowledge acquisition
and experience capture] methods in the construction offices.
They give them (departing veteran employees) a hand shake,
and say "Good Luck."

Another set of interesting comments that were made on several Florida

questionnaires had to do with the use of private construction engineering and inspection firms

known as CEIs. These CEI consultants are utilized for contract administration and


inspection operations on substantial amounts of the highway construction work that is

currently being contracted out by the FDOT. Typically CEI firms in the state of Florida

regularly hire retiring FDOT personnel and resell their services back to the Department.

Although these individuals are no longer technically employed by the FDOT, the Department

still benefits from their knowledge and expertise. Whereas some in the FDOT question

certain aspects of this practice, as evidenced by the following two comments, all agree that

today, the use of CEIs is an integral and established part of highway construction operations

within the state of Florida.

An FDOT Construction Training Engineer has this to say about CEI firms:

A lot of our employees retire with 30+ years of experience
and go to work for a consultant (CEI) that has a contract with
us. Thus we never loose their experience or knowledge, we
just pay them more for it.

A Project Manager with the FDOT made similar comments with respect to CEI firms

and departing veteran practitioners:

This experience is not truly lost because most (90%) of the
departing employees immediately go to work for CEI
consultants who work directly with the Department. The
Department, looses the opportunity to direct these personnel
in ways which would better benefit the people of Florida.

2.1.7 Summary

As previously noted in the research objectives, one of the fundamental purposes of

the KA & EC questionnaire was the identification and prioritization of the general categories

of highway construction work and operations wherein the loss of experience was felt to be

most critical. The area distinguished as most acute would then become the focus of

continued research and development. Analysis of Section II and Section III of the KA & EC


questionnaire indicated that in both the North American and Florida surveys, basically the

same general categories of work and areas of operations were rated as those most affected

by the loss of experience Based on these rather conclusive results, it was determined that

the research effort from this point forward would be concentrated on inspection operations

associated with the construction of new bridges. Although other categories of work and

areas of operations rated at near similar levels, it was felt that this particular selection offered

the best opportunity for the development of a prototype system that would appeal to the

widest audience within the FDOT.

Another observation that can be drawn from a final review of the KA & EC

questionnaire is the apparent lack of any kind of functional implementation of knowledge

based programs among the various responding transportation agencies. The survey was

distributed specifically to those personnel who were in positions of supervising the

construction operations within their agencies. The questionnaire purposely made no direct

references to the term "expert systems" (see Chapter 3 for a discussion of this technology),

in order to ascertain the practical level of use of these types of systems without unduly

prompting such responses. It is very interesting to note that from a total 44 completed

questionnaires, only two respondents made any mention at all of expert systems as a method

by which their department was attempting to capture and disseminate construction

knowledge and experience. One of the two, the New York State Department of

Transportation, actually commented on the fact that after participating in a research project

for the development of an asphalt paving expert system in the early 1990s [Williams et al.,

1990], the department reviewed the findings and decided not to pursue such an approach.

The other agency, however, the Alberta Department of Transportation and Utilities, did


report a significant commitment to the development of expert systems. Since 1990, this

organization has initiated the development of 16 expert systems, of which 11 have been fully

implemented. However, by their own admission, these systems require an inordinate

dedication of departmental resources and time, which has caused somewhat of a reduction

in the popularity of continuing these types of efforts in the future.

2.2 TRB Information on Current Practices

2.2.1 TRB Synthesis on Knowledge Based Expert Systems

During the literature review process, details of which are presented in Chapter 3, a

Transportation Research Board (TRB) report entitled "Knowledge Based Expert Systems

in Transportation, A Synthesis of Highway Practice" was uncovered [Cohn and Harris,

1992]. As part of this synthesis, a survey was conducted in an attempt to ascertain the

current level of development and implementation of knowledge based expert systems among

this country's SHAs. Table 2.9 represents the outcome of this survey. It should be noted

the numbers listed under the column heading "Stage of Development" indicate that the

responding state has been involved with one or more knowledge based expert systems

(KBES) at the designated stage of development as described in the table's legend (1 through

6). The numerical sequence, however, does not necessarily match the activity areas listed in

the last column entitled "KBES Activity Area." Examination of Table 2.9 suggests that the

transportation industry appears to be somewhat more involved in the development of KBES

technology than was evidenced by responses to this dissertation's KA & EC questionnaire.

Table 2.9 Level of KBES Activity Among the SHAs of the United States

Responding Stage of KBES
State Development Activity Areas

California 1, 2, 3, 4, 5 Hazardous materials; Traffic incident
management; Water quality; Concrete
products; KBES priority
Connecticut 3, 5 Pavement rating;
Impact attenuator design
Illinois 1, 3 KBES priority
Emergency response
Kansas 1, 3 Concrete construction;
Concrete pavements
Maryland 3 Freeway incident management
Minnesota 4 Processing truck permits
New Jersey 3, 5 Noise barrier design;
Infrastructure risk management
New York 3, 4, 5, 6 Snow problem location; Asphalt paving
inspection; Pavement marking;
Concrete analysis; Infrastructure risk
management; Steel bridge inspection
Oklahoma 1 KBES state of the art
Oregon 3 Truck weight analysis
Pennsylvania 3, 5 Automated bridge design / drafting;
Structural failure analysis
South Dakota 3 Processing truck permits
Texas 2, 3, 4 Bridge rail retrofit;
Constructability enhancement;
Pavement analysis
Utah 3 Construction evaluation
Virginia 5 Traffic control in work zones;
Disposition of old bridges
LEGEND KBES = Knowledge Based Expert Systems
1= Conceptual; 2 = Prototype in development;
3 = Prototype under testing; 4 = Detailed KBES in development
5 = KBES in use; 6 = Project terminated

Source: [Cohn and Harris 1992]

2.2.2 Survey of the TRB Construction Management Committee

As a follow up to the KA & EC survey and influenced by the TRB synthesis on

knowledge based expert systems in transportation, it was decided that a subsequent letter of

inquiry would be transmitted to a slightly different focus group of industry practitioners, ones

who may have additional information with respect to the more theoretical aspects of

capturing highway construction knowledge and experience. To this end, a directory of names

was compiled from a current list of members of the Construction Management Committee

(A2F05) of the TRB. It was felt that these people represented a more research oriented cross

section of the highway construction industry. However, in keeping with the objectives of

surveying industry personnel specifically, all members of the Committee who were

academicians were eliminated from the list. This left a final total of 18 people from a variety

of different transportation related organizations. The breakdown of their affiliations is as


A) One person was from the Norway Public Roads Administration.

B) One member was employed by the TRB National Research Council.

C) One was from the Federal Highway Administration.

D) Six of the 18 worked for various SHAs.

E) One was from the L.A. County Transportation Authority.

F) Eight were from various private contracting and engineering firms.

Each of these 18 individuals was then sent a letter of inquiry, a generic copy of which

appears in Appendix D along with the associated distribution list. Two representative

examples of the responses received, one being from Parsons Brinckerhoff Construction

Services, Inc., and the other from Martin L Cawley & Associates, is included as Appendix


E. In all, nine out of the original 18 contacted members responded. As was the case with

the original KA & EC survey, expert systems were not specifically mentioned in order to

gauge their level of acceptance among this particular group. Although many of the

comments received were very interesting and well thought out, not a single respondent

referred to the existence of any type of expert systems as a method by which their

organization was attempting to capture construction knowledge and experience.

Another point of agreement between the information gathered through this letter of

inquiry, and that gleaned from the responses to the KA & EC questionnaire, was the

popularity of documenting construction knowledge through the development of construction

manuals. As an example of another such construction manual, Appendix F contains copies

of the cover, the Foreword, the Table of Contents, and subsections I to V of Section 550

(Structural Deck Inspection Guide), as reproduced from the New York State Department of

Transportation's Construction Supervision Manual [New York, 1984]. Although this

publication was received as a result of contact with the New York Department of

Transportation via the letter of inquiry, it should be noted that this manual was also

mentioned in the comments from the New York respondent to the KA & EC questionnaire.

2.3 Current Practices Within U.S. Army Corps of Engineers

2.3.1 General Comments

Although this research effort was focused on highway construction, communication

was established with several large construction organizations that were not specifically

affiliated with the transportation industry. From these preliminary investigations, the U.S.


Army Corps of Engineers clearly set itself apart from other construction entities by the level

of commitment this organization has placed on acquiring and capturing the construction

knowledge and expertise of its personnel.

2.3.2 Jacksonville District Corps of Engineers

Initial contact with the Corps was made through their district office in Jacksonville,

Florida. One outcome of preliminary interviews conducted with members of this office was

the reference to another sample of a construction inspection manual. Appendix G contains

the cover, the Foreword, the Table of Contents, and Sections 2G-01 (General) and 2G-02

(General Requirements) from Chapter 2G (Pile Construction), reproduced from Volume 2

of a four part handbook entitled Construction Inspector's Guide [U.S. Army Corps of

Engineers, 1986]. As was the case with the manuals obtained from the various SHAs, refer

to Appendix C and Appendix F for examples, this document also was written from the

position of managing construction work from the perspective of the government agency

charged with administering the contract.

Another interesting methodology initiated by the Corps in an effort to capture

construction knowledge and experience, is their program of developing "Lessons Learned

Reports" for the analysis of special problems associated with projects that were constructed

under their jurisdiction. An example of such a report appears in Appendix H of this

dissertation. This Lessons Learned Report, generated in the Jacksonville District Corps of

Engineers office, was based on a recently completed project known as the Cerrillos Dam

project. What makes this report especially useful, is the Corps' insistence on relating the

lessons learned from the Cerrillos Dam project to a similar upcoming project, known as the


Portuguese Dam. Not only does this report identify problematic areas encountered during

design, construction, and ongoing operations of a completed project, it also institutes a

procedural method for application of past lessons learned to a specific upcoming project.

What this program represents is a systematic approach, on the part of the Corps, attempting

to capture knowledge and expertise by documenting the experiences of their personnel with

respect to a specific construction project. Furthermore the establishment of a process which

requires that the lessons learned from the Cerrillos Dam project be implemented on the

upcoming Portuguese Dam project, is a very sound method that should help to mitigate the

repetitive occurrence of past problems on future projects.

2.3.3 U.S. Army Construction Engineering Research Laboratories General comments

The examples, as per Appendices G and H, of some of the techniques within the

Corps for capturing knowledge and experience which have been presented up to this point

are certainly worthwhile endeavors which document organizational construction expertise.

The key word here is "document" as it refers to the traditional paper-based methods of

storage and distribution of information. With the advent of the personal computer, and the

proliferation of electronically based information management systems, one would assume that

somewhere within the Corps there must exist more computerized approaches for the capture

and dissemination of construction knowledge and experience.

The U.S. Army Corps of Engineers is the largest public engineering organization in

the world. Falling under their jurisdiction, is the administration of the construction programs

for both the Army and the Air Force, at an annual budget of over five billion dollars, not to


mention their duties associated with managing a myriad of domestic engineering concerns,

such as keeping this nation's waterways navigable [USACERL, 1993a]. In 1969, the Corps

established the U.S. Army Construction Engineering Research Laboratories (USACERL) in

an effort to develop new construction innovations that would serve to enhance the Corps'

future capabilities of managing their growing network of construction and maintenance

related operations. Over the years, USACERL has become one of this country's premier

construction research and development institutions, and as such, it seemed to be a likely place

to continue the search for more state-of-the-art systems that may possibly take advantage of

today's emerging technologies. Developmental knowledge based expert systems

Preliminary discussions were initiated with the USACERL headquarters in

Champaign, Illinois. Results of these conversations led to the discovery of several knowledge

based expert systems (KBES) that were currently under development dealing with a variety

of construction related topics. Fact Sheets, provided by USACERL, describing two such

systems have been included in Appendix I. The first example is that of a KBES designed to

"assist Corps management and technical personnel in using the Design/Build method of

construction contracting" [USACERL, 1993b]. The second Fact sheet also describes an

expert system called Claims Guidance System (CGS). According to the documentation, CGS

analyzes the "relevant information regarding a particular claim" provided to the system by

the user, and based on current legal precedence, generates a set of expert recommendations

[USACERL, 1993c]. ARMS and the BCOE Advisor system

Although expert systems are one method by which USACERL is attempting to

electronically capture construction expertise, the inherent narrow focus, coupled with the

arduous task of creating KBES rule sets has led to research into alternate approaches. One

of the most interesting and comprehensive efforts underway at USACERL is their work on

the programs known as ARMS (Automated Review Management System), and the BCOE

(Biddability, Constructability, Operability, and Environmental compliance) Advisor system,

which is a developmental extension of ARMS.

In an initial step towards automating the design review process, the computer experts

at USACERL began developing ARMS. This program is basically an extensive database of

project review comments maintained by the Technical Center of Expertise (TCX) at the

Sacramento District, Corps of Engineers office in California. The fundamental purpose of

ARMS is to provide all members of a project team a "management tool for the collection,

resolution, and storage of comments generated during the design/construction life of a

project." Figure 2.3 presents a schematic flowchart which illustrates the method by which

ARMS manages the comments which arise as a result of the design review process. Quoting

again from the ARMS manual, "This program is tailored to replace the current system of

receiving and resolving hand written design comments" [U.S. Army Corps of Engineers,


As a part of the overall design review process, that which ARMS was created to help

manage, the U.S. Army Corps of Engineers requires what is known as a BCOE (Biddability,

Constructability, Operability, and Environmental compliance) review on all of their projects.

The concept is to involve construction personnel in this BCOE review in order to identify the



Source: [Roessler et al., 1993]

Figure 2.3 Schematic Flowchart of ARMS Operations




- - - -


construction related problems. Upon recognition of these problems, the designer is then

informed, and thus can modify the design so as to avoid costly construction contract

modifications and minimize the cost of building operations and maintenance. As a

computerized extension of ARMS, in the realm ofBCOE reviews, USACERL initiated work

on a program called the BCOE Advisor. It is interesting to note that the developers of this

system originally experimented with a rule based KBES upon which to build their BCOE

Advisor prototype, however, this approach soon was found to be inappropriate for the

unique nature of the construction projects being reviewed. Instead, it was decided that the

BCOE Advisor would be produced on rBASE, a commercially available relational database

software package marketed by the Microrim Corporation. According to the BCOE Advisor

programmers, rBASE was selected because at the time it was the only PC platform database

system that supported the American National Standard Institute's (ANSI) Standard Query

Language (SQL) [Roessler et al., 1993]. Storing the BCOE Advisor data in SQL would

then make it possible to import and export comments to and from ARMS directly in a

standardized format.

What the developers of the BCOE Advisor system attempted to do was to augment

the BCOE process by creating a database system that could 1) conceptually access past

design review comments from ARMS, 2) modify these comments with respect to the current

project being reviewed, and 3) store these new modified comments for future use. Figure 2.4

illustrates the basic operations of the BCOE Advisor system and its interface capabilities with

ARMS. Fundamentally, as Roessler et al. [1993] suggest, what this established was a system

that provided a "lessons learned capturing program to assist in the generation of high quality,

ARMS compatible, design review comments."


e c


2.4 Summary of the Survey of Current Practices

2.4.1 The General Category of Work for Further Concentration

One of the fundamental purposes of surveying current practices within the highway

construction industry was to distinguish that area of highway construction work in which

those in the industry felt that the effects associated with the loss of experience due to the

departure of veteran personnel was most acute. From the responses to Section II of the KA

& EC questionnaire as presented in this chapter, the general category of "New Bridge

Construction" clearly established itself as the highest rated category in level of importance

given to experience based issues.

The survey also looked at those operative and administrative duties that typical

construction staff members are most regularly involved in. Again from the results of the KA

& EC questionnaire, the three areas from Section III that were identified as being most

affected by the loss of experience were "Constructability Analysis," "Inspection Operations,"

and "Quality Control." The area of "Constructability Analysis," although obviously related

to construction operations, is also heavily involved in the design aspects of highway work.

As such, and keeping in mind that the original scope of the study was specifically on that of

construction operations, it was decided that this area would not be singled out for further

concentration. With respect to construction operations, one of the primary functions of

"Quality Control" is accomplished by ensuring that the finished product has been built in

accordances with the plans and specifications, as well as all other applicable construction

related practices and requirements. Therefore, much of the burden of monitoring the field

quality and compliance of the finished product falls squarely on the shoulders of those people

who are in charge of the "Inspection Operations."


Taking into account the totality of the KA & EC questionnaire responses, coupled

with the original goals as set forth in the research objectives, it was decided that the

subsequent knowledge acquisition efforts were going to be limited to "New Bridge

Construction" with a focused attention on "Inspection Operations." Not only did the survey

bear out these results, but conversation with many FDOT personnel indicated that this

concentrated effort would reach the widest audience within the Department. And as has been

noted, since the end-user in this case was to be the FDOT, it only served to strengthen this


2.4.2 Knowledge Acquisition and Experience Capture Methods Inspection and operational manuals

By far, the most popular technique of capturing construction expertise was the

utilization of written construction inspection and operational manuals. Although these types

of manuals were found to exist in most of the construction organizations contacted, many of

the comments received regarding these documents acknowledged their limited effectiveness.

It is not that these manuals do not contain significant amounts of quality information, rather

it is more a function of the cumbersome way in which this information is presented. U.S. Army Corps of Engineers Lessons Learned reports

The U.S. Army Corps of Engineers Lessons Learned reporting program was a

method of experience capture that appeared to be very productive. The concept of

documenting the problematic areas of a particular job creates a more systematic approach to

knowledge acquisition. The fact that project personnel are able to discuss relatively recent

occurrences ofa specific nature, yield comments that are more focused and ultimately more


useful. This technique was deemed to be one which demonstrated promise for

implementation with respect to this dissertation's efforts in regards to the capture of highway

construction knowledge and experience. Current computerized developmental efforts

With respect to cutting-edge computer based technologies, the field of expert systems

seems to be the one that has attracted the most attention of late. Although there are a

number of transportation related KBES programs currently under development, as evidenced

by Table 2.9, results of the KA & EC questionnaire, along with the subsequent contacts made

with other organizations, both in and out of the highway construction industry, indicate that

the functional utilization of expert systems by those personnel who are directly associated

with day-to-day construction operations is very limited or nonexistent. Wentworth [1993]

suggests that an explanation for this lack of practical acceptance by the industry may be due

to the fact that in his opinion, highway applications of expert systems "appear to be more

developer-driven than user-demanded."

Another interesting computer application uncovered at the U.S. Army Construction

Engineering Research Laboratories (USACERL) was their conceptual database for the

management of comments generated through the BCOE (Biddability, Constructability,

Operability and Environmental compliance) portion of the design review process. The

program, which is called the BCOE Advisor is a very innovative method of storage and

retrieval of text based comments. Although not specifically related to the highway

construction industry, the idea of relating comments of similar subject matter and providing

conceptual access to this information certainly is an approach worth investigating.


One class of information software that has yet to be discussed but which shows great

promise for managing construction related text and graphics is the technology known as

hypertext. Williams [1991], in an article describing a hypertext asphalt paving system that

was developed in conjunction with the New York State Department of Transportation

(NYSDOT), defines hypertext as a "database system of text and graphics that allows a

reader to jump from idea to idea depending on one's interest." This article in particular was

chosen to quote because Williams is the same person who had participated in the previously

mentioned development of the asphalt paving expert system [Williams et al., 1990] that was

subsequently abandoned by the NYSDOT. According to the NYSDOT respondent to the

KA & EC questionnaire, after departmental review of the hypertext system as compared to

the expert system, the hypertext system was deemed to be more suited to their needs and is

currently being successfully utilized.

Another example of a hypertext application found within the transportation industry

is a program called the Highway Constructability Improvement System (HCIS), which was

developed for the Washington State Department of Transportation (WSDOT). In an article

written for the Transportation Research Board, HCIS is described as database of information

extracted from five years worth of WSDOT change orders. The WSDOT felt that by using

HCIS, engineers at their design office could access knowledge from past construction related

experiences that resulted in change orders, and hopefully avoid similar errors in preparing

future design plans and specifications [Lee et al., 1991].

2.5 Final Comments on the Survey of Current Practices

In general, based on the survey of current practices, there did not appear to be any

comprehensive highway construction knowledge acquisition and experience capture

programs that had gained any significant levels of acceptance among those practitioners who

are intimately involved in the day-to-day operations of building this country's highway

systems. It was felt that for a knowledge acquisition and experience capture program to be

successful in an organization such as the FDOT, the information delivery system must cater

to the needs of the end user. Although some promising developments in the area of

information management were uncovered, further review into the current literature associated

with this field of study is required, and as such, will be pursued in detail in Chapter 3 of this



3.1 Introduction

Up to this point in the research endeavor, significant effort had been focused on the

identification of the methods by which different agencies within the highway construction

industry were attempting to acquire knowledge and experience, and disseminate this captured

knowledge to other members of the organization. Although no one singular system that fully

addressed all the issues of capturing construction knowledge and experience as set forth in

this dissertation's research objectives had been uncovered, in particular, the three information

management technologies of expert systems, hypertext, and database management systems

had emerged as likely candidates for utilization in one form or another as potential tools for

possible realization of the stated objectives. It was apparent that as stand alone entities,

none of these three branches of information management could be considered as fully

responsive to the needs of the proposed system. However, it was felt that by integrating

certain aspects of each type of computer software, a prototype computerized system could

be developed that would create an intuitive, user-friendly environment for the capture and

dissemination of highway construction knowledge and experience.

With this in mind and concurrent with the survey of current practices, as detailed in

Chapter 2, an extensive literature survey was undertaken utilizing the University of Florida's


on-line searching capabilities of the Library User Information Services (LUIS). Additionally,

a computer database search was conducted through the Southern Technology Applications

Center (STAC), which like the university, is also located in Gainesville, Florida. At STAC,

two of the most comprehensive, commercially available engineering index services, DIALOG

(File 63--TRIS) and COMPENDEX, were queried. These searches were focused on the

distinct fields of study relating to expert systems, hypertext, and database management

systems, in order to gain a better understanding of the various capabilities of each of these

information management techniques. With a firm grasp of the underlying functionality

associated with these technologies, intelligent decisions could then be made with respect to

the level and strategies of integration that would be pursued. The results of this review

process, along with conclusions for the proposed system integration, will be presented in the

following sections of this chapter.

3.2 Knowledge Based Expert Systems

3.2.1 General Comments

Although the standard knowledge based expert systems (KBES) that are currently

being developed in the highway construction industry have shown themselves to be inherently

narrowly focused and typically not well received by industry practitioners, it was felt that

there may be certain properties associated with expert systems that might turn out to be very

useful. In order to determine what aspects of the KBES approach may be worthwhile in the

context of this research effort, the concept of what an expert system is and exactly what it

does will be explored next.

3.2.2 Historical Background

Knowledge based expert systems (KBES), as they exist today, are a direct outgrowth

of the artificial intelligence techniques that began to develop after World War II. In 1956,

a group of scientists from such fields as electrical engineering, mathematics, neurology and,

psychology, got together at Dartmouth College in New Hampshire to discuss the possibilities

of utilizing the computer as a means of simulating various aspects of human intelligence. The

proposed intent of the Dartmouth Conference was to explore the conceptual supposition

"that every aspect of learning or any other feature of intelligence can in principle be so

precisely described, that a machine can be made to simulate it." They termed this new

technology Artificial Intelligence (AI) [Rose, 1984].

One result of the Dartmouth Conference was the establishment of future aspirations

for the AI field. It was forecasted that by 1970, a computer would be able to do the


1) be a grandmaster at chess;

2) discover significant new mathematical theorems;

3) understand spoken languages, and provide language translations; and

4) compose music of classical quality.

By the mid 1960s, it had become painfully apparent that these lofty goals of true artificial

intelligence set by the Dartmouth Conference were not going to be met, and in hindsight they

were very unrealistic. The AI community regrouped and began to consider more modest

goals for the intelligent machine. They agreed that knowledge was the essential ingredient

of intelligence. They also realized that the computer, despite its sizeable capacity for data

storage, was not able to store and process the incredible amount of information that would


be necessary to simulate actual cognitive human intelligence. They therefore decided that for

the time being, they would focus their research and adopt the following strategies:

1) be more modest;

2) be more focused; and

3) direct system development towards a narrow sector (domain) of
expertise, rather than attempt to simulate general overall human

The name given to this new subfield ofAI was Expert Systems or Knowledge Based Expert

Systems (KBES) [Ignizio, 1989]. Figure 3.1 illustrates the history of expert systems as they

evolved from artificial intelligence [Harmon and King, 1985].

3.2.3 Generalized Overview of Knowledge Based Expert Systems Definition

A KBES is a sophisticated computer program that manipulates knowledge of a

specific domain in such a way as to solve complex problems that would otherwise require

extensive human expertise [Waterman, 1986; Rolston, 1988]. Probably one of the most

frequently quoted and comprehensive definitions of what constitutes a KBES can be

attributed to Professor Edward Feigenbaum of Stanford University, a leading authority in

expert systems research. Feigenbaum defines an expert system as follows [Harmon and King,


An expert system is an intelligent computer program that
uses knowledge and inference procedures to solve problems
that are difficult enough to require significant human expertise
for their solution. Knowledge necessary to perform at such
a level, plus the inference procedures used, can be thought of
as a model of the expertise of the best practitioners of the

.o c

0 E

3.5 2\

The knowledge of an expert system consists of facts and
heuristics. The 'facts' constitute a body of information that
is widely shared, publicly available, and generally agreed upon
by the experts in the field. The 'heuristics' are mostly private,
little-discussed rules of good judgement (rules of plausible
reasoning, rules of good guessing) that characterize expert-
level decision making in the field. The performance level of
an expert system is primarily a function of the size and the
quality of the knowledge base it possesses [p. 5]. Functional components of a generic KBES General comments. The typical architecture of a generic KBES is

illustrated in Figure 3.2 [Ignizio, 1991]. Based on this figure, a brief discussion of each

component and its functional relationship to the overall system will be presented next. The human/computer interface. The dotted horizontal line represents the

cut off point between human users and the computer operations. Below the line there is the

"User" who is the non-expert person utilizing the system. The "Knowledge Engineer" is the

person who interfaces with the domain expert, defines the expert's knowledge, and models

it in such a way so as it can be loaded into the computer. The process by which the

"Knowledge Engineer" seeks out and captures this knowledge and expertise is commonly

referred to as knowledge acquisition [Parsaye and Chignell, 1988]. Knowledge acquisition

has developed into a subspecialty in its own right, and will be discussed in more detail in

Chapter 4. Continuing with the human/computer interrelationship, the "Interface" is the

system's component that controls all input/output functions that take place between the

computer and either the "User" or the "Knowledge Engineer." The knowledge base. The "Knowledge Base" is universally recognized

as the "heart and soul" of any KBES [Parsaye and Chignell, 1988; Ignizio, 1991]. As earlier

stated in Feigenbaum's eloquent description of an expert system, the "Knowledge Base"








I 0


stores two types of knowledge (facts and rules). To reiterate, facts are statements whose

validity are widely accepted as truth. Facts are obviously significant in assuring the accuracy

of the system, but they alone cannot be used for reasoning. By relating facts together with

rules however, relationships can be represented, reasoning can then be inferred, and new facts

can be derived. Representation of the knowledge in the "Knowledge Base" can be achieved

utilizing a variety of methods which include production (IF...THEN...) rules, semantic

networks, object-attribute-value (OAV) triplets, and frames [Waterman, 1986; Adeli, 1988;

Dym and Levitt, 1991]. Of these knowledge representation schemes, the production rule

approach is by far the most widely used and easiest to understand. With this in mind, any

future reference to knowledge representation within this generalized overview presentation,

will concentrate on the production rule metaphor. The working memory. The "Working Memory," which is often referred

to as the context component of the KBES, is similar to the "Knowledge Base" in that it also

contains facts. However, the difference is that the facts within the "Knowledge Base" are

statically imbedded, that is to say that these facts are existing and do not undergo change

during system utilization. The "Working Memory" on the other hand, dynamically stores

new facts which are generated by the system itself in one of two ways. "Working Memory"

facts are either derived from the cycling of the "Inference Engine," or they are produced as

a result of consultations with the "User." The inference engine. The "Inference Engine" is the mechanism by which

the KBES locates existing knowledge and infers new knowledge from the "Knowledge

Base." The "Inference Engine" accomplishes two main objectives:

1) It examines the existing facts and rules within the "Knowledge Base," and
when possible it adds new facts to the "Working Memory."


2) It also controls the order in which the inferences are made. The most
common inference strategies are forward chaining, backward chaining,
or some type of combination of these two approaches. Forward chaining. As noted previously, in a rule based KBES, the

knowledge is represented by a collection of IF... THEN... production rules. The concept of

forward chaining, also known as "bottom up" or "data driven" searching, can best be

explained by examining what happens within the generic KBES that has been under

discussion. In general terms, initial data is supplied to the KBES through consultation with

the "User." This data is then compared to the IF portion of the rules in the "Knowledge

Base." When a particular IF part of a rule is deemed to be true, that is to say that the KBES

matches supplied data to an IF condition, then the rule "fires," creating a new fact (the THEN

portion of the rule) which is immediately added to the "Working Memory." In an iterative

process, the rule base is reexamined continuing to utilize the initial supplied data in

conjunction with the new inferred facts in an attempt to deduce a solution to the problem at

hand. Forward chaining is therefore best suited for situations wherein the KBES is called on

to interpret a set of incoming facts, and reach some kind of conclusion based on this incoming

data [Maher, 1987]. Backward chaining. Backward chaining, which is also commonly referred

to as "top down" or "goal driven" searching, is a much more difficult strategy to understand,

but in simplistic terms, it can be thought of as basically the reverse of forward chaining.

Under the backward chaining approach, the KBES compares the desired goal (hypothesis)

to the THEN portion of the production rules in an attempt to evaluate whether or not the IF

part of the applicable rule or rules can be justified. If successful, the goal is established and

the KBES reports its results, otherwise another hypothesis is formed and the "Inference


Engine" repeats the procedure [Bielwaski and Lewand, 1991]. Backward chaining, due to

the fact that the inference strategy is goal driven rather than data driven, would therefore tend

to be more useful under those conditions where the number of possible solutions is limited. The rule adjuster. The final module that will be discussed herein is the

"Rule Adjuster." This component is really nothing more than an editor for the rules. In other

words, the "Rule Adjuster" is the tool by which the "Knowledge Engineer" enters and

modifies the rules of the "Knowledge Base" during the KBES development and subsequent


3.3 Hypertext

3.3.1 General Statement

Further research into the standard KBES approach for capturing and disseminating

highway construction knowledge and experience revealed, as was preliminarily suspected,

that the basic architecture of such a system represented a rather restrictive developmental

environment. The effort required to capture the vast amounts of construction related

knowledge and expertise possessed by a large organization, such as the FDOT, would be

extremely cost prohibitive, and as such not very practical with respect to the stated objectives

of this research endeavor. As previously noted in the problem statement contained in

Chapter 1, and further confirmed by the results of the KA & EC questionnaire presented in

Chapter 2, most of the SHAs in the United States have invested significant dollars and time

over the years developing an assortment of construction related documents that are meant

to assist their personnel in supervising highway construction operations conducted within

their particular jurisdictions. Although it is generally agreed upon that these various


published materials contain considerable amounts of useful construction knowledge, most

industry practitioners acknowledge that timely and effective access of this information is

often a significant problem. The emerging technology of hypertext represents a very

practical solution to this predicament.

Hypertext, in generic terms, is a nonlinear information management system that

allows the user to access information in a more natural way, similar to the way in which that

user might store and access information in his or her own mind. This less impeded

methodology for accessing information creates a more free flowing environment that enables

the user to explore the knowledge base driven more by his or her own interests, rather than

by the predefined structure inherent in traditional paper based linear documents. Similar to

the discussion of the KBES approach, the balance of this section on hypertext will present

a historical background of hypertext, as well as a general overview of this rapidly developing

field of computerized technology.

3.3.2 Historical Background

The origin of the hypertext concept is universally attributed to Dr. Vannevar Bush,

who among his many accomplishments, served as the Director of the Office of Scientific

Research and Development under President Franklin Delano Roosevelt during World War

II. In 1945, Bush published an article in The Atlantic Monthly in which he described a purely

theoretical device which he called the memex, short for memory extender [Bush, 1945].

Although he did not specifically use the term "hypertext" in any of his writings about the

memex, all of the experts in the field of hypertext development agree that Bush's imaginative

memex device was the non-computerized forerunner of all of today's computer hypertext



In his now famous article, entitled "As We May Think," Bush wrote of his concerns

about the post-war explosion of scientific information which would make it nearly impossible

for the research specialists of the day to follow all the new developments associated with a

particular field of study. Today this situation is geometrically worse, but even in 1945 Bush

realized the need to enable people to access information more effectively than was possible

via traditional paper based documentation. He envisioned the memex device, which he

explains as follows [Bush, 1945]:

A memex is a device in which an individual stores his books,
records, and communications, and which can be mechanized
so that it may be consulted with exceeding speed and
flexibility. It is an enlarged intimate supplement to his memory
[pp. 106-107].

The mechanism that Bush goes on to describe is an ingenious machine that would be

capable of storing millions upon millions of pages of written material reduced onto microfilm.

By inputting the code of a particular document into the memex, via a keyboard, the user

would instantly be able to view the document in question. Furthermore, the memex provided

the capability of creating links between various pages of a single document, as well as the

ability to access pages from other completely separate sources. This linking of items is

described by Bush [1945] as:

... associative indexing, the basic idea of which is a provision
by which any item may be caused at will to select immediately
and automatically another. This is the essential feature of the
memex. The process of tying two items together (by
association) is the important thing [p. 107].

Not much was done in the field of hypertext research until the early 1960s, when a

young electrical engineer by the name of Douglas C. Engelbart from the Stanford Research

Institute, influenced by Bush's article of 1945, began work on a similar vision in which he


saw computers as a means of assisting thought, or as he referred to it, "the augmentation of

the human intellect" [Nelson, 1992]. In an article, entitled "A Conceptual Framework for the

Augmentation of Man's Intellect," Engelbart [1963] writes of his beliefs that the computer

represented a new stage in human development.

In this stage, the symbols with which the human represents
the concepts he is manipulating can be arranged before his
eyes, moved, stored, recalled, operated upon according to
extremely complex rules--all in very rapid response to a
minimum amount of information supplied by the human, by
means of special cooperative technological devices. In the
limit of what we now imagine, this could be a computer, with
which individuals could communicate rapidly and easily,
coupled to a three dimensional color display with which
extremely sophisticated images could be constructed [p. 14].

Engelbart's ideas, as presented in his 1963 article, led to the development, in 1968,

of a system which he named NLS (oN Line System). NLS according to Engelbart, was an

experimental tool designed to aid his research group in their efforts by [Engelbart and

English, 1968]:

... placing in computer store all of our specifications, plans,
designs, programs, documentation, reports, memos,
bibliography and reference notes, etc., and doing all of our
scratch work, planning, designing, debugging, etc., and a
good deal of our intercommunication, via the consoles

These consoles were very sophisticated by the standards of the late 1960s and included

television imaging, as well as a variety of input devices, the most famous of which is known

today as a "mouse" [Conklin, 1987].

Concurrent with Engelbart's development of NLS, which has evolved over the years

and is now called Augment, another hypertext pioneer by the name of Ted Nelson began

work on his own personal concept of "augmentation," emphasizing "the creation of a literary

environment on a global scale." In 1965 Nelson coined the term "hypertext" in describing


the nonlinear nature of text based storage and retrieval represented by his conceptual "Project

Xanadu" [Conklin, 1987; Parsaye et al., 1989].

Some ofNelson's early efforts on Project Xanadu were accomplished while he was

affiliated with Brown University in the mid to late 1960s. Although Project Xanadu has only

recently begun to find limited commercial applications through its sale in 1988 to Autodesk,

Inc., a large software development company, the work conducted by Nelson while at Brown

University directly influenced the development of the world's first computerized working

hypertext system. A colleague of Nelson's, a man by the name of Andries van Dam, is

generally given credit for heading up the research group that unveiled this hypertext system

in 1967. This system, which was called "The Hypertext Editing System," ran in a 128K

memory partition of a small IBM System 360 mainframe computer and was funded by an

IBM research contract. Upon completion of the project, the system was sold by IBM to the

Houston Manned Spacecraft Center, where it was subsequently used to produce a variety of

documentation for the Apollo space missions [Conklin, 1987; Nielsen, 1990].

For the better part of the next twenty years, work continued on the development of

a number of hypertext systems, however, with the exception of very limited commercial

applications, these programs compromised in-house endeavors utilized only by the

institutions where the systems were originally designed. By the early 1980s commercial

versions of some of these research oriented projects did begin to come to the general

marketplace. These early hypertext systems, such as NoteCards, developed by the computer

scientists at Xerox PARC (Palo Alto Research Center), were designed to run on workstations

[Berk and Devlin, 1991]. The requirement of workstations was due to the fact that at this

time, personal computers, although in existence, had not yet developed enough internal

power to run such systems.


The first mass marketed, personal computer based, hypertext system that achieved

any level of commercial popularity was a program known as Guide. Guide, which began as

a research project at the University of Kent at Canterbury in 1982 [Parsaye et al., 1989], was

introduced in 1986 by a software company called OWL (Office Workstations Limited).

Originally, Guide only ran on Macintosh computers, but shortly after its release in 1986, a

version that would run on IBM compatible machines under the Windows operating system

was developed [White, 1992]. However, it was not until the release of Hypercard by Apple

in 1987 that the concept of hypertext truly became mainstream.

Although an adequate programing platform in its own right, the real impetus for

Hypercard's surge to the forefront of the hypertext industry was a decision by Apple to

bundle Hypercard, free of charge, into the operating system of every Macintosh computer

sold after 1987. What this has done obviously, is to ensure that every Macintosh user who

purchased their machines after 1987 has access to Hypercard whether or not they initially

showed any interest in the software. Apple's visionary marketing approach has led to

Hypercard becoming by far the most widely used hypertext system to date, claiming, as of

1991, a world wide user base of literally millions [Bielawski and Lewand, 1991;Woodhead,


3.3.3 Generalized Overview of Hypertext Definitions Hypertext. Any discussion of the term "hypertext" would be somewhat

remiss if it did not include the observations of Ted Nelson, the man who originally invented

the word. From one his early publications on the subject, Nelson [1967] suggests the

following definition:

Hypertext is the combination of natural-language text with the
computer's capacity for interactive branching or dynamic
display, when explicitly used as a medium. Or, to define it
more broadly, "hypertext" is the generic term for any text
which cannot be printed (or printed conveniently) on a
conventional page, or used conveniently when bound between
conventional covers. "Nonlinear text" might be a fair
approximation [p. 13].

Probably the easiest way to explain what hypertext is, is to contrast it with traditional text.

Traditional text, whether paper based or electronic, is sequential in nature, having a linear

structure defining the order by which the document is intended to be read. Hypertext, on the

other hand, is non-sequential, and therefore can allow the reader to explore the document in

any order he or she chooses, driven more by personal interest than document structure. Hypermedia. Back when Nelson first began using the term hypertext, he

was basically describing plain-text electronic documents. However in today's multimedia

landscape, an electronic document has taken on a much wider definition. More than just

conventional text, contemporary computerized documents can also contain graphics

(drawings and pictures), animation, audio, video, as well as multitasking references to other

computer programing routines external to the particular document being viewed. Therefore,

considering this expanded version of what constitutes an electronic document, some of the

leading experts in the field of hypertext research, prefer to use the term "hypermedia" as a

means of highlighting the multimedia aspects of their developmental systems. Figure 3.3

presents an illustration representing this idea of hypermedia as being a fusion of hypertext

plus multimedia [Howell, 1992].

HYPERtext + MultiMEDIA


Source: [Howell, 1992]

Figure 3.3 Hypermedia as a Fusion of Hypertext and Multimedia "Hypertext"--selected as the generic preference. Whether one calls it

hypertext or hypermedia, the theory is the same, that being the construction of a nonlinear

network of linked pieces of information which are presented in such a fashion as to enable

a user to navigate through this network, accessing desired information in a more natural and

associative manner. Given that there does not appear to be any overwhelming necessity to

distinguish between these two terms, it has been decided that the convention that will be

adopted for this dissertation will be that of utilizing these terms rather interchangeably, with

preference given to the more traditional terminology of "hypertext." The basic concept of hypertext

The fundamental concept underlying hypertext is rather simple. Information is

organized or "chunked" into relatively small, self-contained "nodes" which are connected via

"links." Figure 3.4 [Bubbers and Christian, 1992] serves to illustrate this idea, while

simultaneously presenting some of the historical background of hypertext as previously

discussed. As can be noted from this figure, there are three separate nodes connected by four

associative links. The various words surrounded by boxes, for example the names of


3:E ~~ ~ *~0ri
4- 2 00 0

04 C* i *3

= o co:
Sc4 ) 4-

*~ m 6) Z
0 C0
z > :J Cl. 0a c



Q o

0 0
ca cau


.- .- 'o- Ei 4
0>o E E A)
+ 0 0 0
4) E ~ P f
> "0 c ("t -0 4
c, d
0 cO- E p L
r,. t) (7N

Q~~j z >~ : r
w co w 0s .p
1. s. M +~


"Vannevar Bush" and "Douglas Engelbart" in Node 1, are known as "hot keys," "link

anchors" or "points." In virtually every contemporary, mouse driven hypertext system, when

the mouse is dragged across the hot keys, which are typically delineated from the rest of the

text by being displayed in a different color, the cursor changes from it standard shape (usually

an arrow head) to a special shape (often a closed right hand with an extended index finger).

At this point, the user only has to click the mouse, causing the system to jump automatically

via the link to the node that correlates to whichever hot key was clicked on.

Again looking at Figure 3.4, assume the user is located in Node 1 and is reading the

information about hypertext. If for example, the user determines that he or she would like

more information associated with the highlighted hot key of "Vannevar Bush," a simple click

of the mouse will cause Node 3 to immediately pop up for viewing. Although Figure 3.4

illustrates a case involving only plain-text based nodes, utilizing the new generation of

multimedia hypertext systems, a skilled developer could have programed the system to access

for example, a picture of Vannevar Bush, an audio recording of his voice, or maybe a film

clip, if one existed. Referring back to Dr. Bush's landmark article, "As We May Think"

[Bush, 1945], published over 50 years ago, it is truly amazing to note just how similar

today's hypertext systems are to his imaginative memex device.

As previously noted in the discussion of expert systems presented earlier in this

chapter, one of the standard methods for representing knowledge within the knowledge base

of a generic KBES was the utilization of semantic networks, also known as semantic nets.

Waterman [1986] describes these structures as a collection of points, which are called

"nodes," connected by various links, which are commonly termed "arcs" in the semantic

network vernacular. Figure 3.5, reproduced from Jeff Conklin's definitive article,



-------" ------------ --------)

.. .. -- -- -





"Hypertext: An Introduction and Survey" [Conklin, 1987], illustrates what he describes as

the method by which a typical hypertext system establishes "correspondence between

windows and links in the display, and nodes and links in the database." From this figure it

can be seen quite readily that conceptually speaking, the notion of hypertext is intimately

related to the idea of semantic nets. This being the case, it is apparent that hypertext is more

than simply an innovative word processing paradigm. Rather hypertext is by all accounts a

knowledge representation tool in its own right, capable of storing and representing

knowledge both in the nodes themselves, and through the associative linkage structure that

connects these discrete nodes.

3.4 Database Management Systems

3.4.1 General Comments

As was established in the previously presented review of hypertext systems, one of

the fundamental and most appealing features associated with hypertext is the free flowing

environment this class of software provides for exploration and navigation through a

particular base of knowledge and information. However, this unstructured landscape can

often lead to a user experiencing a feeling of disorientation, commonly referred to as being

"lost in hyperspace." One possible method of overcoming this predicament is to incorporate

certain aspects of modern database management systems (DBMS), as a means of providing

structure to the inherently unstructured world of hypertext.

The following sections will present a closer look at the historical background

associated with the emergence and evolution of contemporary DBMS. Additionally, the

three most prominent models of data structuring will be discussed, with special attention


being paid to the relational model, since it is this model which has come dominate today's

personal computing DBMS software packages.

3.4.2 Historical Background

The history of DBMS dates back to the early 1960s, when a number of individual

corporations in the United States began to produce programs that were created in order to

solve in-house data related problems specifically encountered by these particular companies.

Probably the most famous of these early efforts was a system developed in 1962 by the

General Electric Company (GEC), which was called the Integrated Data Store (IDS)

[Beynon-Davies, 1991]. Several years later B.F. Goodrich expressed significant interest in

the IDS package, however IDS had been designed to only run on the GEC brand of

mainframe computers. Since these computers were not compatible with B.F. Goodrich's

IBM (International Business Machines Corporation) systems, they decided to rewrite IDS

so that it would operate on the newly released IBM System 360 family of computers. Soon

after beginning work on translating IDS, B.F. Goodrich entered into a marketing agreement

with a man by the name of John Cullinane, and together they launched the IDMS DBMS

which became one of the dominant DBMS for IBM mainframes throughout the 1970s and

1980s [Brodie and Manola, 1989; Beynon-Davies, 1991].

Development of IDS and the subsequent release of IDMS represented the maturation

of the network model of DBMS. The network model, however, was only one of three basic

models of data structuring that were evolving somewhat simultaneously. Another of these

data structuring techniques being researched during the 1960s was the hierarchal model. In

1965, in response to the massive information handling requirements associated with the

Apollo moon program, North American, which later became Rockwell International


Corporation, and IBM co-developed a hierarchal model DBMS that eventually was released

by IBM in 1970 under the name of IMS (Information Management System) [Cardenas, 1985;

Parsaye et al., 1989].

At about the same time that IBM was developing their IMS DBMS, another IBM

computer scientist by the name of Dr. E. F. Codd, located at the IBM Research Laboratory

in San Jose, California, began work on a general purpose programming language based on

set theory and logic, which he called relational programming [Brodie and Manola, 1989].

In 1970, Codd published his landmark article, "A Relational Model of Data for Large Shared

Data Banks" [Codd, 1970], which established the relational model on which all subsequently

developed relational DBMS were to be based. The relational model did not initially meet

with wide spread acceptance, and by the mid 1970s the DBMS landscape had become

dominated by the other two models of the network and hierarchal data structuring


Although originally not very popular, relational modeling gradually did become more

recognized as a legitimate structure for DBMS, and by 1976 IBM, through its research center

in San Jose, California, was able to develop System R [Astrahan et al., 1976], which became

the first working relational DBMS for mainframe computers. Another prominent

experimental mainframe relational DBMS released around the same time as System R, was

a program called INGRESS [Stonebraker et al., 1976], which was developed at the

University of California, Berkeley.

The relational model of DBMS continued its existence almost exclusively within the

bounds of university and other research institute settings, until 1983, at which time IBM

unveiled DB2, their first commercially released relational DBMS for mainframe computers,


which was a direct outgrowth of their earlier experimental work with System R [Salzberg,

1986]. Approximately at the same point in time, a software company by the name of Ashton-

Tate released dBASE II, which went on to become the dominant DBMS for the newly

emerging personal computing market. According to Brodie and Manola [1989], by 1988

over 2.7 million copies of the dBASE relational DBMS software package for personal

computers had been sold.

3.4.3 Generalized Overview of Database Management Systems Definitions General comments. To accurately describe exactly what is a database

management system (DBMS), a number of database terms and their usage will be presented

first, followed by a working definition of a DBMS. Data. Data, independently speaking, are nothing more than a collection

of facts and figures, that by themselves lack any real significance. All data within a database

can be broken down into two main categories, namely alphanumeric data and numeric data

[Date, 1990]. Alphanumeric data consists of alphabetic characters (the letters A through Z)

and numerical characters (the numbers 0 through 9), as well as a variety of specialized

symbols such as the pound sign (#) and the dollar sign ($), to name two. Numeric data, on

the other hand, are strictly a set of numeric digits that can be quantified. Although when

stored in a database, both alphanumeric and numeric data represent information, these two

classifications take on different roles in their applications. The numeric data within a

database are used as numbers in computational operations, while alphanumeric data can only

be used as strings of text for identification and labeling purposes.

69 Fields, records, and files. In a database, the smallest unit of data, whether

alphanumeric or numeric, is commonly referred to as either a "field," a "data item" or an

"attribute." A collection of these "fields" constitutes a logical "record," also known as an

"entity." A "file" is an assortment of occurrences of the same "record" types [Cardenas,

1985]. Database. A database can be described as a bank of "records" stored in

"files" interrelated by a means of specific relationships. A database is basically a repository

for stored data which is both integrated and shared [Date, 1990]. All database systems can

be characterized by their efforts to achieve the following four properties [Beynon -Davies,


1) Data Independence--Due to the concept of shared data, the data in a
database must be independent of the storage structure and access

2) Data Integration--Again because of sharing capabilities, a database
should contain as little duplicated or unused data as possible.

3) Data Integrity--Given that numerous applications are intending to interact
with a particular database, it is extremely important that the data must be
maintained at a high level of consistency and accuracy [Cardenas, 1985].

4) Separate Logical and Physical Views--What this implies is that a database
systems should attempt to separate the end-user's view of the data from
the data's physical computerized representation. Database management system. In its most generic form, a database

management system (DBMS) is a system capable of supporting and managing an integrated

database. This suggests, that basically a DBMS is a family of software applications which

have been developed to act as an interface between the end-user and all system interactions

with the database. Data modeling structures for contemporary database management systems General comments. The fundamental component of any DBMS is the

method by which the data within the database is organized and structured. In the commercial

world of DBMS, the marketplace is still dominated by the three traditional data modeling

techniques of the network model, the hierarchal model, and the relational model, with the

relational model having almost exclusively captured the personal computing market. It

should be noted at this point, that there does exist a number of other data modeling

techniques such as semantic models, entity-relationship diagrams, and most notably object-

oriented designs [Stonebraker, 1988 Martire and Nuttall, 1993], which recently have

experienced significant research attention [Parsaye and Chignell, 1993]. In fact currently

there are several commercially available object-oriented DBMS software packages available,

the first and probably most established of which is a program called GemStone, marketed by

a company by the name of Servologic [Beynon-Davies, 1991; Parsaye and Chignell, 1993].

However, for the purposes of the following discussion on data modeling, only the three

prominent techniques (network, hierarchal and relational) will be examined further. The hierarchal model. The hierarchal model, sometimes referred to as the

file system, is the oldest and most rigid of the three standard database modeling techniques

[Date, 1990]. Figure 3.6 [Chou, 1985] illustrates a typical hierarchal representation of the

different relationships among the data fields associated with the instructors, the classes, and

the students. The relationships represented in this figure are limited to strictly one-to-one

associations. For example, the class "Business 101" is directly related to the instructor "Peter

Roberts" on a one-to-one basis. This connection can also be defined in terms of what is

known as a parent-child relationship [Martire and Nuttall, 1993]. Referring again to the












figure, and in terms of the parent-child metaphor, the instructor "Peter Roberts" is said to be

the parent of the class "Business 101," which is the child. The network model. One of the inherent disadvantages associated with

the hierarchal model is that of data redundancy, which is a result of this model's strict one-to-

one architectural structure. As an example of this, notice that in Figure 3.6, due to the rigid

structure, three of the student data items ("James Smith," "Eileen Hason," and "Danny

Walter") must be appear twice in order to satisfy the model. The network model of data

structuring was developed in large part to address this problem [Taylor, 1989]. Figure 3.7

[Chou, 1985] depicts the same database as that of Figure 3.6, except for the fact that in

Figure 3.7, the data is structured based on the network model. Examination of this figure

demonstrates that in fact this model does eliminate data redundancy, in other words, in the

network model, every data item is unique. However this aversion of data redundancy does

come with a price, namely a much more complex linkage structure. The relational model in general. In both the hierarchal and network

models, relationships between data items are strictly controlled by their explicit links, or

record instances as they are sometimes referred to [Woodhead, 1991]. These links are

represented in Figures 3.6 and 3.7 as the solid lines connecting the various data items. This

static feature creates a situation wherein navigation through and/or manipulation of the data,

not to mention the problems associated with adding, deleting, or modifying a record, require

the services of a very skillful database programmer. These difficulties are all but eliminated

under the relational model, due to the fact that the data is structured in terms of a two

dimensional tabular matrix consisting of a set of named columns, or fields, and an arbitrary

number of rows, also known as records. This highly intuitive structure, which is simply a




2 o\
os o
0 X


E) o

0 (
o .


S\ E
c aS

0 000
(0 l--i
m g~

___ w~I


common table, enables a non-sophisticated user to rather easily set up and manipulate

considerable amounts of data in a very interactive and natural fashion [Chou, 1985;

Woodhead, 1991]. The relational model as illustrated in Figure 3.8. Figure 3.8 represents the

same set of data items as presented in Figures 3.6 and 3.7, except this database is structured

in terms of the relational model. The relational table in Figure 3.8 is organized into nine

rows, which are also called tuples [Beynon-Davies, 1991] or records, and three columns,

each corresponding to a distinct field of data, which in this case are labeled Instructor, Class,

and Student. As each data record is entered, it is sequentially and automatically given an

arbitrary number, which for this relational table would be a number between 1 and 9,

depending on the order of entry. Each of the nine records is assigned a single corresponding

attribute within each of the three data fields of Instructor, Class, and Student. This simple

tabular structure therefore serves as a methodology of creating nine uniquely identifiable

database records. Parsaye et al. [1989] rightfully note that it is this "theoretical purity" and

close relationship to natural logic that more than any other factor has propelled the relational

model to the forefront of today's DBMS software packages, especially in the realm of

personal computers where the users often tend to be less sophisticated.

3.4.4 A Closer Look at Relational Database Management Systems General comments

Since the proposed information management prototype system intended for this

dissertation was to be developed under the IBM compatible personal computing

environment, and given that, as has previously been noted, relational DBMS have evolved

as the dominant force on this platform, logic dictated the selection of this model for further

6. 0 0 o >

SI > >0
SC0 0 0

0 00 0

, u -9, 0) 0
S0 0 0 0-
S0 E E E Ec 0

U CD D CD C. C C C )
00 m m)m Lu w L I

O0 O O O 0 0 0 0
r 0 0 0 0 0 0 00
a:S a: a: C a: a: a
C 0 0 0 0 a E C
St o o o & B f E c
o a .S a )) 0 ) 0(0 C (V

C m C lt L ) O w r- o m)



examination. Presented next will be a closer look at some of the key points associated with

managing data in a relational DBMS. Data manipulation within relational databases Operating on relations. One of the distinguishing characteristics of the

relational model, as opposed to the hierarchal or network model, is that in relational

databases, the data manipulation is designed to operate on entire files (relational tables) of

data, rather than on individual records or fields within a file. The term manipulation, as used

here, refers to the types of operations that a user can perform on data stored in a relational

database. This manipulation of the relational model consists of a set of operators collectively

known as the relational algebra [Beynon-Davies, 1991]. The relational algebra. The relational model basically consists of a list of

relations and their associated attributes [Date, 1990]. Data retrieval within this model is

accomplished by using the relational algebra as a means of manipulating these relationships.

Parsaye and Chignell [1988] specify the five fundamental operations in the relational algebra

as follows:

1) Selection--Selects certain rows from a table.

2) Projection--Removes certain columns from a table.

3) Product--Multiplies two tables together.

4) Union-Adds two tables together.

5) Difference--Subtracts one table from another table. Structured Query Language. With the evolution of the relational model,

came the development of higher-order languages designed to provide access to the relational

model and extract (retrieve) different data sets depending on the specified request of the user.


Of these data access languages, or commonly referred to as query languages, one in

particular, IBM's Structured Query Language (SQL), has come to be accepted as the

dominant approach for relational query languages. In fact, in 1986, SQL was adopted by the

American National Standards Institute (ANSI) as the official industry standard [Fleming and

von Halle, 1989]. An example of a generic SQL command. As an example of the most

fundamental SQL instruction, Date [1989] suggests a generic sample of the SQL command

(query) "SELECT" taking the following form:

select < Attribute,, Attribute2 ... Attribute >

from < Relation,, Relation2 ... Relation >

where < Condition >

In terms of the relational algebra, from which SQL is a direct descendant [Chorafas, 1989],

the SQL "SELECT" command is made up the select portion of the query clause which is

equivalent to the relational algebra operation of projection. The from segment of the

"SELECT' statement matches the relational algebra operation of product, while the relational

algebra counterpart of where is the operation of select.

3.5 Summary and Conclusions

3.5.1 General Comments

Having effected a comprehensive background study of the three technologies of

expert systems, hypertext, and database management systems, the next step was to determine

what aspects of each were useful for integration into the proposed prototype information

management system. The following sections will summarize these observations, focusing on


the aspects deemed worthwhile based on the stated needs of the end user and the anticipated

industry sector within which this system will be intended to function.

3.5.2 Considerations Regarding Proposed Integrated Environment General programing requirements

Bielawski and Lewand [1991], co-authors of Intelligent Systems Design Integrating

Expert Systems. Hypermedia, and Database Technologies, recognized as one of the definitive

books in this newly emerging field of study, suggest that the power of today's IBM

compatible and Macintosh personal computers, coupled with the myriad of available

developmental tools or "shells," make this platform ideal for the development of integrated

computer software systems. This idea, along with the fact that the design of any prototype

system should be based on the needs of the end user, led to the selection of the IBM

compatible personal computer windows operating system, given that this is the system of

choice for practically all of the intended end users, who in the case of this research project,

are FDOT construction personnel.

Another point that should be emphasized, is that given that the nature of this research

endeavor was more geared towards developing a conceptual systematic approach, rather than

undertaking an exercise in conventional computer programming, it was determined that

higher levels of programing paradigms should be investigated and utilized whenever possible.

This notion will be revisited towards the end of this chapter under the discussion regarding

software requirements for the prototype system. Intended knowledge base content as determining factor for hypertext underpinnings

Although this subject will be covered more thoroughly in Chapter 4, a preliminary


understanding of the structural makeup of the intended knowledge base would be required

in order to effectively identify which aspects of which technologies were to be utilized. As

has previously been mentioned, most of this country's SHAs have over the years amassed a

considerable base of construction related documentation, which in essence captures much of

the knowledge and expertise possessed by these particular organizations and their personnel.

In an effort to utilize the significant investment represented by these documents, it seemed

only prudent that the prototype system take advantage of this wealth of information. With

this in mind, and considering each of the three technologies analyzed, clearly the proposed

approach should utilize hypertext as the backbone of the developmental philosophy for

managing this naturally occurring text based bank of construction knowledge and experience. Integrating database strategies for added structure

Nanard et al. [1993], conceptually describe hypertext as fundamentally a relational

database with unlimited and unrestricted links between the records, fields, and files. This

metaphor is highly appropriate to this dissertation's intention of employing database

strategies as a method of providing structure to the inherently unstructured environment of

a pure hypertext system. From the field of database management, two basic modeling

strategies were to be incorporated. The first, was a hierarchal structure based on the

traditional hierarchal model as presented earlier in this chapter. The idea here was to embed

a one-to-one, hard-wired, parent-child, relational linkage scheme throughout the entire

architecture of the hypertexed system of nodes. Details of how this was accomplished will

be examined further in Chapter 4.

Inspiration for the other database structuring strategy came from two sources, one

of which, the BCOE Advisor system [Roessler et al., 1993], was detailed in Chapter 2. This


system, developed by USACERL, utilizes the commercially available rBASE relational

DBMS, along with SQL querying capabilities, as a methodology for storing and retrieving

conceptually indexed design review comments. The second basis of influence was a paper,

uncovered as a result of this review of published literature, written by Dario Lucarella [1990].

In his article, entitled "A Model for Hypertext-Based Information Retrieval," Lucarella

presents an interesting discussion with respect to the difference between browsing and

searching. He describes browsing as characterized by the user knowing "where he is in a

network, and wanting to know what information exist at that location." While when

searching the user "presumably knows what he wants, and wishes to know where in the

database it is." Figure 3.9 illustrates Lucarella's proposed integrated tactic which provides

capabilities for both browsing and searching.

Borrowing from these two innovative approaches, the second of the database

strategies was formulated in an attempt to utilize the relational model of database

management as a method of indexing each node in the prototype hypertext network. In

theory, a node could be conceptually linked to a set of attributes based on the informational

content that it contained. These nodes could then be stored in a relational table enabling a

user to directly query the system in order to access a group of hypertext nodes, specifically

related to his or her area of interest, regardless of the nodes physical location within the

hypertext network. In other words, a user would not have to blindly explore the network

via the hard-wired links, node to node, in search of the desired information. Rather he or she

could retrieve all nodes within the system having to do with a particular subject, and then

continue navigational browsing from that point forward. Details of the implementation of

this strategy will be presented as part of Chapter 5.

0 0
u Z


F9 P= O 3: cn c4 Z ~3


^ I-
/ u



) J
-< ib

3 I-

i\ Utilization of a KBES inference engine for dynamic linking

Assuming that the relational database structuring strategy was attainable, the other

somewhat more ambitious proposal was to supplement this static structure with a form of

dynamic linking. To reiterate, each table would be established as a relational structure

compromised of individual records, each of which corresponding to a single node in the

network, conceptually organized in a "controlled indexed" [Carlson, 1989] format. Once

this was accomplished, it was envisioned that the system could recognize the particular node

where the user was located, search the entire network for all other nodes of related subject

matter, and then return a list for the user to browse. The determination of "related" nodes

could be accomplished utilizing an embedded generic rule set coupled with a forward

chaining inference strategy. Presentation of the fruition of this strategy is also included in

Chapter 5.

3.5.3 Software Requirements for Prototype System

Based on the discussion thus far, four general requirements of any potential software

development tool have been established as follows:

1) The software must operate under the IBM compatible personal computing
windows environment.

2) The software package must have rich and fundamentally underlying
hypertext capabilities.

3) The software package must support the relational database model and
preferable provide SQL capabilities for querying a relational database.
As noted previously, the benefit of SQL is that it has been accepted as the
industry standard relational database query language.

4) The software package must be capable of delivering an imbedded KBES
inference engine that can employ the forward chaining strategy.

One other item that was previously mentioned earlier in this section, and which will

be expanded on at this point, is the concept of fully integrated, higher-level developmental

programming tools. As was noted, one of the expressed objectives of the research's software

development efforts was the exploitation of commercially available packages, rather than

attempting to develop the prototype system using established conventional programming

languages such as C++, Pascal, LISP or PROLOG. Although these languages are very

powerful, they require inordinate amounts of programming time in relation to functional

return on this manpower investment. Under the guise of a construction research project, it

is far more important to identify the cutting edge computer technologies and attempt to

implement them with respect to construction related issues.

Along this same line of reasoning, it is also highly beneficial to utilize only one

vendor's software package rather then attempt to integrate distinct and separate packages

developed by different manufacturers. Although all systems that operate under the windows

environment theoretically inherit some basic level of platform wide integrative capabilities,

typically, experience shows that the dependence of fully seamless integration across software

vendors is usually not a recommended practice. The basis of this discussion on the

importance of maintaining a single developmental environment, led to the fifth and final

requirement for the potential software development tool, which is as follows:

5) If possible, all functional requirements established in points 1) through 4)
as listed above, should be accomplished by the use of a single vendor's
higher-level windows developmental programing software package.

3.5.4 Final Selection of Software Package for Prototype System

With the established requirements of the potential software package in mind, some

of the newest and commercially available developmental programming tools were evaluated.

This original review was accomplished via preliminary searches of appropriate articles found

in such computer trade publications as Infoworld, PC Magazine, and Byte. Results of this

preliminary review yielded the two candidates as listed below for final consideration for

selection as the software package to be utilized:

1) The Intelligence Compiler (I/C)--distributed by IntellegenceWare, Inc.,
whose president is Kamran Parsaye, principle author of three of the books
uncovered during this literature review and referenced throughout this
dissertation [Parsaye and Chignell, 1988; Parsaye et al., 1989; Parsaye
and Chignell, 1993].

2) KnowledgePro Windows (KPWin)--distributed by Knowledge Garden

Both of these companies were contacted and from each, a set of demonstration diskettes and

a standard manufacture's information package was received. Appendix J contains a copy of

the general product sheets associated with these two highly integrated and powerful software

packages. Upon completion of the review of the demonstration diskettes and the associated

product literature, the final choice was made to proceed with the KPWin software package,

namely because it was felt that this programming tool supported a fuller hypertext

environment as compared to that of (I/C).

3.5.5 Final Comments

With the literature review complete and the software package selected, the focus of

this dissertation would now be shifted from evaluating the efforts of others to initiating work

specific to accomplishing the remaining stated objectives of this research endeavor.

Presented next in Chapter 4 will be the issues involved in developing the focused base of

highway construction knowledge and experience. Chapter 5 will follow with a

comprehensive explanation of the development and subsequent testing of the computerized

information delivery system, which from this point on will be referred to as the IN REACH

system, an acronym for Intelligent iNformation Retrieval and Expert Advice in the

Construction of Highways.


4.1 Introduction

As was noted during the earlier discussions regarding knowledge based expert

systems (KBES), the task of capturing the knowledge and experience of the domain experts,

commonly referred to as knowledge acquisition [Parsaye and Chignell, 1988], is universally

recognized as the critical phase in the production of any successful knowledge based system

[Waterman, 1986; McGraw and Harbison-Briggs, 1989; Dym and Levitt, 1991].

Fiegenbaum's assessment of the late 1970s, in which he noted that at that time, knowledge

acquisition was the "bottleneck" in expert systems development [Fiegenbaum, 1977], still

holds true today. This chapter will present a brief look at the traditional approach as

compared to this dissertation's modified strategy, with respect to knowledge acquisition.

Additionally, this chapter will examine the foundational knowledge base developed for the

IN REACH system, as well as include a discussion of the documents utilized along with a

presentation of the embedded hierarchal model within the underlying hypertext system.

4.2 The Traditional Approach to Knowledge Acquisition

4.2.1 General Comments

With the proliferation of KBES developmental efforts over the last three decades, the

concept of knowledge acquisition has developed into its own separate field of study.

Evidence of this can be seen by the fact that today there exist numerous texts [Kidd, 1987;

McGraw and Harbison-Briggs, 1989; Adeli, 1990] completely dedicated to this subject.

Furthermore, reviews of any of the standard text books on KBES [Parsaye and Chignell,

1988; Dym and Levitt, 1991; Ignizio, 1991] indicate that invariably there are at least one or

two chapters on this critical area of expert system development. This representative list of

published books does not include the myriad of articles [Cohn et al., 1988; De La Garza et

al., 1988; Hanna et al., 1992] that have been published over the last several years in the

various technical engineering journals. Presented next will be a brief overview of some

aspects of the traditional approach to knowledge acquisition.

4.2.2 An Overview of the Traditional Approach

According to McGraw and Harbison-Briggs [1989, p. 8], knowledge acquisition can

be thought of as a process that encompasses "both 1) the task of reducing an exhaustive body

of diverse domain knowledge into a precise, easily modifiable set of facts and rules; and 2)

the tools and methods that support the system development." Hanna et al. [1992] also take

this approach of describing knowledge acquisition as involving the entire operation of

constructing a knowledge base. They go on to concisely define what they perceive as the

three basic stages of "extracting knowledge and creating a knowledge base" as follows: