Using case-based reasoning for knowledge management

Material Information

Using case-based reasoning for knowledge management filling the void left by rule-based expert systems
Austin, Jeffery S
Place of Publication:
Denver, CO
University of Colorado Denver
Publication Date:
Physical Description:
ix, 106 leaves : ; 28 cm.


Subjects / Keywords:
Knowledge management ( lcsh )
Case-based reasoning ( lcsh )
Expert systems (Computer science) ( lcsh )
Rule-based programming ( lcsh )
Case-based reasoning ( fast )
Expert systems (Computer science) ( fast )
Knowledge management ( fast )
Rule-based programming ( fast )
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )


Thesis (M.S.)--University of Colorado at Denver, 1999. Computer science
Includes bibliographical references (leaves 88-106).
General Note:
Department of Computer Science and Engineering
Statement of Responsibility:
by Jeffery S. Austin.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
42611426 ( OCLC )


This item is only available as the following downloads:

Full Text


USING CASE-BASED REASONING FOR KNOWLEDGE MANAGEMENT: FILLING THE VOID LEFT BY RULE-BASED EXPERT SYSTEMS by Jeffery S. Austin B.S., University of Alabama in Huntsville, 1992 A thesis submitted to the University of Colorado at Denver in partial fulfillment of the requirements for the degree of Master of Science Computer Science 1999


This thesis for the Master of Science degree by Jeffery S. Austin has been approved by Chris Smith Date


Austin, Jeffery S. (M.S., Computer Science) Using Case-Based Reasoning for Knowledge Management: Filling the Void Left by Rule-Based Expert Systems Thesis directed by Associate Professor William Wolfe ABSTRACT Industry'calls the retention, distribution, and reuse of organizational know-how knowledge management. A key tool for knowledge management is the creation of an automated "corporate memory". This thesis provides an overview of case-based reasoning (CBR) and investigates the possible use of CBR corporate memory systems to help fill the void rule-based expert systems currently leave in certain domain types. The thesis further examines the potential for CBR systems to serve as a knowledge management tool by creating corporate memory in all cases where: the domain is not well defined or numerous exceptions exist; there is some experience within the domain; there are few experts; and versus computers, humans can still more efficiently adapt past experience iii


(cases) to solve a class or type of problem within a domain. Finally, the thesis demonstrates the findings using an example: a CBR corporate memory system for the Base Realignment and Closure (BRAC) process. This abstract accurately represents the content of the candidate's thesis. I recommend its publication. iv


CONTENTS Figures . . . . . . . . . . . . . . . . . . . . ix CHAPTER 1 INTRODUCTION . . . . . . . . . . . . . . . . 1 Motivation . . . . . . . . . . . . . . . . 1 Problem Overview . . . . . . . . . . . . . 2 Related Work . . . . . . . . . . . . . . . 3 2. CASE BASED REASONING (CBR) OVERVIEW ......... 6 Introduction . . . . . . . . . . . . . . . 6 CBR History & Approach ..................... 7 CBR History . . . . . . . . . . . . . 7 The CBR Approach . . . . . . . . . . . 8 Advantages/Disadvantages of CBR ............ 9 Cases & Representation .................... 10 Storage (Memory Organization) 13 v


Indexing . . . . . . . . . . . . . . . . . 13 Retrieval . . . . . . . . . . . . . . . . . 15 Adaptation . . . . . . . . . . . . . . . . 16 Learning . . . . . . . . . . . . . . . . . 19 CBR System Design, Maintenance, & Testing .. 19 CBR System Design . . . . . . . . . . 19 Maintaining the Case Base ............. 21 I Testing ............................... 21 3. RULE-BASED EXPERT SYSTEMS, CBR, AND KNOWLEDGE MANAGEMENT ........................ 23 Rule-Based Expert Systems .................. 23 Case-Based Reasoning . . . . . . . . . . . 24 Knowledge Management ....................... 25 Corporate Memory: A Renewed Focus for CBR .................... 27 CBR Knowledge Management Systems ........... 31 Criteria for Applying CBR to Knowledge Management ....................... 32 Domain Type . . . . . . . . . . . . . 3 3 User Adaptation . . . . . . . . . . . 33 Domain Experts . . . . . . . . . . . . 3 4 vi


Domain Experience ..................... 34 4. BRAC ADVISOR: AN EXAMPLE CBR CORPORATE MEMORY SYSTEM ..................... 36 The Problem . . . . . . . . . . . . . . . . 3 6 A Proposed CBR Solution .................... 39 Criteria for Applying a CBR Corporate Memory ........................... 40 Domain Type . . . . . . . . . . . . . 41 User Adaptation ....................... 42 Domain Experts ........... . . . . . . 42 Domain Experience ..................... 43 General Advantages of Using CBR for this Problem . . . . . . . . . . . . . 43 BRAC Advisor System Design ................. 45 Choosing a CBR Tool ................... 45 CASPIAN ............................... 46 CASPIAN & BRAC Advisor ................ 49 5. CONCLUSIONS ................................. 54 Case-Based Reasoning . . . . . . . . . . . 54 CBR for Knowledge Management ............... 55 vii


APPENDIX A. BRAC ADVISOR CASE-BASE AND CONTROL DATA ................................ 57 B. SAMPLE RUNS OF BRAC ADVISOR ................. 66 C. BRAC PROCESS OVERVIEW ....................... 80 D. OTHER IMPORTANT BRAC PROBLEM AREAS .......... 83 ANNOTATED BIBLIOGRAPHY ............................... 88 viii


FIGURES Figure 3-1. BRAC Advisor Cases: Categories of Property ................................ 52 3-2. BRAC Advisor Cases: Categories of Reuse .... 53 ix


CHAPTER 1 INTRODUCTION Motivation I was introduced to the concept of case-based reasoning during my graduate level studies in computer science at the University of Colorado at Denver. While attending the university, I was also serving as an Army logistics officer at a military base that was identified for closure under the Base Real-ignment and Closure (BRAC) Acts of 1995. I immediately began to wonder about the benefits that case-based reasoning could provide the Army, and in particular, my work with BRAC. The greatest challenge I saw my organization facing was conducting a base closure, seemingly without the benefit of experience gained from rounds o_f base closure that had already occurred. It seemed almost every task undertaken was something that was being done for the very first time when this was clearly not the case. It was for these reasons that I chose to learn more about case-based 1


reasoning and to explore whether or not case-based reasoning could prove useful as a knowledge sharing tool. Problem Overview Expert systems are created to help the users' performance approach that of a human expert for a specific task or within a certain problem domain. However, traditional expert systems are not well suited to domains that are poorly defined or where there are numerous exceptions. Rather than leaving a void in these domains, case-based reasoning advisor systems may be created. While case-based reasoning advisor systems cannot guarantee an optimum solution as rule-based expert systems sometimes can, case-based reasoning advisor systems are at least a solution that can help users improve their performance in these difficult domains. In these domains where rule-based expert systems are not practical, case-based reasoning advisor systems can be created to serve as a "corporate memory", ,allowing users to reuse solutions to past, similar problems to solve current problems. 2


Related Work The only previous work I found specifically related to using case-based reasoning systems for corporate memories was the paper, "The Experience-Sharing Architecture: A Case-Study in Corporate-Wide Case-Based Software Quality Control" by Hiroaki Kitano and Hideo Shimazu (Kitano 1996) In this paper, the authors discuss the merits of using a CBR system as part of a corporatewide strategic information system to enhance the sharing of experience as they have done at NEC through the creation of the.CBR corporate memory system, SQUAD. However, the focus of the paper is Case-Method, a methodology for the development of CBR systems. I was pleased to find that the authors agree that too much of the CBR community's attention is placed on automated adaptation, rather than being spent on creating more CBR advisor systems. It was Ian Watson's work, Applying CaseBased Reasoning: Techniques for Enterprise Systems (Watson 1997), that provided the domain characterization for where CBR advisor systems are most applicable. Using this criteria of sorts for when CBR advisors are useful, I propose that CBR advisor systems can (and should) be created in all domains that meet this criteria. I argue that by doing so, CBR advisor systems, serving as a 3


corporate memory, can help fill the void expert systems leave in certain domains. The sources I found most useful in developing an understanding of case-based reasoning were Janet Kolodner's Case-Based Reasoning (Kolodner 1993), Ian Watson's Applying Case-Based Reasoning: Techniques for EntekPrise Systems (Watson 1997), and Agnar Aamodt's and Enric Plaza's paper, "Case-Based Reasoning: Foundational Issues, Methodological Variations, and System Approaches" (Aamodt 1994). The former work and David Leake's collection of papers, Case-Based Reasoning: Experiences. Lessons. and Future Directions (Leake 1996), provide indepth and cutting edge information on case-based reasoning. For a broad, but thorough look at traditional expert systems, I found Chris Nikolopoulos' Expert Systems: Introduction to First and Second Generation and Hybrid Knowledge Based Systems (Nikolopoulos 1997) most useful. Finally, for information on expert systems in the corporate world, corporate memories, and knowledge management, I found the following to be most beneficial: Kriz's Knowledge-Based Expert Systems in Industry (Kriz 1986), Otto Kuhn's and Andreas Abecker's paper, "Corporate Memories for Knowledge Management in Industrial Practice: Prospects and Challenges" (Kuhn 1998), and Thomas Koulopoulos' "Knowledge Management: 4


Toward Creating the 'Knowing Enterprise"' (Koulopoulos 1997) 5


CHAPTER 2 CASE-BASED REASONING OVERVIEW Introduction The purpose of my thesis is to demonstrate that CBR can help fill the void rule-based expert systems currently leave in domains that are poorly understood or where there are many exceptions to rules simply by creating corporate memory for use as a knowledge management tool. Moreover, my goal is also to identify the criteria for when CBR may be successfully implemented as a corporate memory solution. My thesis continues with a overview of CBR in this chapter. The next chapter discusses using CBR for knowledge management. In chapter three, I demonstrate my claims about using CBR as a knowledge management tool through an example, proposing a CBR corporate memory system for the BRAC process. Moreover, I develop a demonstration prototype (BRAC Advisor) The thesis ends with my conclusions in chapter four. 6


CBR History & Approach CBR History Case-based reasoning has a relatively short history, beginning with the role of reminding in humans as found in the cognitive science research of Roger Schank's group at Yale University. Janet Kolodner, who was part of Schank's group, developed the first CBR system, CYRUS, in 1983 which was based upon Schank's dynamic memory model (Kolodner 1984) Case-based reasoning moved from the world of pure academics to the commercial world in the late eighties when Cognitive Systems, through Federal government sponsorship, created what carne to be ReMind, the first CBR tool (Watson 1997). Today, there are hundreds of CBR systems designed to perform or assist with diagnosis, planning, legal reasoning, design, arbitration, tutoring, advising, and a host of other functions. 7


The CBR Approach Case-Based Reasoning is the successful recalling of solutions to past problems (both successes and failures) for use in solving a current problem. "The CBR approach is based on two tenants about the nature of the world. The first tenant is that the world is regular: similar problems have similar solutions. Consequently, solutions for similar prior problems are a useful starting point for new problem-solving. The second tenant is that the types of problems an agent encounters tend to recur. Consequently, future problems are likely to be similar to current problems (Leake 1996) ." Take, for example, making a sandwich. At some point in my life, I had to learn the steps necessary to make a peanut butter and jelly sandwich. I may have even sought an optimum solution, spending a relatively great amount of time taking small increments of peanut butter or jelly and carefully spreading it on the bread to get the perfect amount I desired with good, even coverage. At some point, this stopped. I no longer wanted or needed to spend time or great thought in making a peanut butter and jelly sandwich. In the familiarity of my home, it seems I can mindlessly gather the ingredients and utensils and have a perfectly delicious sandwich within a couple of minutes. 8


In fact, I give little thought to spreading any familiar substance (mayonnaise, mustard, tuna salad, etc) on bread. When assessing the amount of any substance to load onto the knife for spreading on the bread, I don't stop and calculate how much I need. I just recall about how much I like from my past experience and use about that much again. That is case-based reasoning. It involves a cycle sometimes referred to as the four REs (Watson 1993) : 1. REtrieve the most similar case(s) 2. REuse the case(s) to attempt to solve the problem 3. REvise the proposed solution if necessary 4. REtain the new solution as part of a new case Advantages/Disadvantages of CBR Case-based reasoning has many strengths, but certainly has its share of weaknesses as well. Probably one of the best qualities going for CBR is that it is easily understood because it models what we humans do: reason in current situations based on our past successes and failures in similar situations. Not only is CBR something we are all familiar with, it also often saves time. For example, when we successfully do a task for the first time, we may have to struggle a bit and do some 9


reasoning based on piecing together many facts. However, when confronted with the exact same task again, we recall as much as we can about what we did last time and begin more detailed reasoning from there, saving effort and time. Even when we are not faced with the exact task again, often it is similar enough to a previously encountered task that we can recall the solution to the similar task, modify the past solution, and come up with a successful solution to the new task. This holds true in the realm of expert systems as well. With tasks or domains that are poorly understood or that have many exceptions, it is possible to succeed fairly often by reasoning from past experiences within the problem. domain. The biggest disadvantage of CBR is that, since CBR is most often applied in domains that are poorly understood or where there are many exceptions, CBR systems often are not fully automated. Still, CBR systems are proving very useful whether fully automated or not. Cases & Representation A case is a "contextualized piece of knowledge representing an experience that teaches a lesson fundamental to achieving the goals of the reasoner (Leake 1996) ." So, not every experience is a case, just those 10


that teach one or more lessons important to the domain of the reasoner. The issue of how to represent cases in a case-based reasoning system's case library is a very important matter that continues to receive much research effort. Obviously, the specific purpose of a CBR system will play a critical role in determining what the components of a case should be. However, there are several questions that should be answered regarding case representation in any CBR system (Kolodner 1993): What is a case? What are the component parts of a case? What kinds of knowledge does a case need to encode? What representational formalisms and methodologies are most useful in representing cases? So, in addition to knowing what a case is (its purpose), one has to consider what the component parts will be. Generally, a case has at least three parts: a problem or situation description, a solution or course of action, and the results or outcome. The kinds of knowledge contained in these components include such things as: problem goals and constraints, justifications or expectations for the solution, alternative plans that were assessed to be good or bad, and how the solution 11


could be improved, or how a failure could have been avoided (Kolodner 1993). One must consider how cases will be implemented. For example, will frames, semantic nets, rules, relational database techniques, or a combination of knowledge representations be used? The decisions about what information will be included in a case and how it will be represented has a direct impact on a CBR system's ability to retrieve, rank, compare, perform adaptation, and make repairs. Still, there are other important representation issues. Most CBR systems require human interaction and so representation must take into account human needs as well. Also, cases that are very large, involve a large time slice, or are continuous pose special problems. For very large cases, cases can be represented as a whole with a scheme for locating pieces within the case by indexing the case with its own indexes as well as indexes for its pieces. Alternatively, the pieces could be represented as cases with links to sibling pieces so the larger case may still be manipulated (Kolodner 1993). When a large time slice is involved, the problem description may evolve. Most often, the final problem description is recorded in the case, but sometimes multiple cases are created to represent the significant differences. For continuous environments, determining case boundaries can be very difficult. This is an area 12


where more research is needed. Another area of research is in methods for representing knowledge in multimedia formats. Storage (MemokY Organization) The issue of how to organize and store cases in a CBR system's memory is also an important one. Cases typically represent large chunks of contextual knowledge. The challenge is to balance maintaining the richness of these cases verses the need to have efficient access and retrieval of relevant cases. There are a several methods, called case-memory models, that provide guidance in this area. Academically, the most influential models are the dynamic memory model of Schank and Kolodner, and the category-exemplar model of Porter and Bareiss. However, the most common practice commercially is to use flat file data structures or use conventional relational database structures (Watson 1997). Indexing Indexing helps the CBR system speed up case retrieval, but the trade off is the danger that using 13


indexes may overly restrict what cases are recalled. Often cases provide contextual information of value only to the user, so not all information in a case needs to be indexed. In general, two kinds' of indexes are useful: differences and anomalies (Kolodner 1993). Kolodner provides the example of the Legal Seafoods Restaurant where customers order, pay, eat, and then tip. If such a restaurant was associated with a case, then the fact that Legal Seafoods serves excellent fish chowder may be indexed since it makes it different from other seafood restaurants that only serve clam chowder. The fact that one pays before being served is somewhat of an anomaly, and so would be a good index. When actually selecting an index, a generally accepted set of guidelines is to make sure the index: is predictive, addresses the purposes for which the case will be used, is abstract enough to allow future expansion of the use of the case base, and is concrete enough to be recognized in the future (Watson 1996) Research on indexing continues with focus on vocabularies, automating index selection and creation, searching based on indexes, and organizing cases based on indexes. 14


Retrieval Second to automating the adaptation of past solutions to solve current problems, the retrieval of similarly useful cases is the most challenging area of CBR. A system's ability to successfully retrieve cases depends on many factors, including case and storage organization, situation assessment, indexing and vocabulary, and retrieval, matching, and ranking methods. "Situation assessment is the process of analyzing a raw situation and elaborating it such that its description is in the same vocabulary as cases already in the case library (Kolodner 1993) ." In other words, situation assessment is the process of determining what the indexes for a current problem situation would be if it was already in the case-base. This is a challenging process in itself to successfully accomplish, but it pays great dividends for the retrieval process when implemented. After any situation assessment, the following typically occurs in the retrieval process. Retrieval algorithms use the current problem case and its indexes to search the case base. The retrieval algorithms or heuristics call on matching procedures to assess the degree of match between the current problem case and cases in the case base. The retrieval algorithm returns a list of partially matching cases. Ranking procedures then refine the list of 15


partially matching cases into a smaller list of cases with the most potential for useful similarity. Nearest neighbor and inductive retrieval are the two most used retrieval techniques (Kolodner 1993). The nearest neighbor technique allows the use of background knowledge in the form of relative importance, or weightings, on case attributes. However, this approach's retrieval speed is a weakness due to required comparisons to every case in the case-base, but there are partial solutions to this problem. Inductive Retrieval is extremely quick and performance decreases slowly as the case base size increases. The major disadvantage of inductive retrieval is that its ability to retrieve the most usefully similar cases suffers when those cases contain missing or unknown (fuzzy) data, but there are ways of working around this problem too (Watson 1997) Adaptation "Adaptation looks for prominent differences between the retrieved case and the current problem and then applies formulas or rules that take those differences into account when suggesting a final solution (Watson 1997) ." Automating the adaptation of past solutions to solve current problems is the most challenging area of 16


CBR and receives a tremendous amount of CBR research attention. However, it is not the most important aspect of CBR. I would have to say that the issues associated with improving case retrieval are of greatest importance. This view of adaptation not being so important is in agreement with Watson who states, "While adaptation is useful in many situations, it is by no means essential. Unless adaptation can be done easily using simple and well-understood parameter adjustments or reinstantiations, my advice would be to avoid it (Watson 1997) ." The problem with automating adaptation is that it usually requires rules of some sort. This poses a problem because the most common reason CBR is used rather than a rule-based or other knowledge based system is that the domain is one in which there is little understanding or there are many exceptions. In addition to being too difficult to implement, there are other times when automating adaptation should be avoided. Automation may be inappropriate when preferences are important, value judgments or aesthetic judgments are needed, or because problems are very large (Kolodner 1996) In these cases, adaptation may be best left to humans. If adaptation will be automated in a CBR system, there are generally four major decisions that must be made (Kolodner 1996) : what needs fixing 17


identifying what part of the faulty solution should be changed to carry out the fix identifying applicable adaptation methods and/or heuristics selecting an adaptation strategy and carrying it out Leake also acknowledges that often the adaptation process is best left to the user, but when automating adaptation is important, he suggests the burden of adaptation can be alleviated some by refining other CBR system components. He suggests that: indexes can be refined to favor more adaptable cases; similarity judgments can be based on adaptability; the entire retrieval process can be based directly on a cases adaptability; preparing for adaptation can be done at case storage time; the CBR system can learn from user adaptation by tracing and recording the user's adaptation process for reuse (Leake 1997). Similar to Leake's last suggestion is research in which CBR systems use CBR to perform adaptation. When adaptation is required, "CBR is used to see if a similar adaptation has been done in the past. If so, the adaptation is retrieved and reused (or even possibly revised itself) If [a similar case of successful adaptation is not found] a person is asked to perform the adaptation. Once the adaptation has been done 18


successfully, either by the program or the person, it is stored for future use (Watson 1993) ." Learning Case-base reasoning systems learn by adding new cases to their case base. However, the cases that are added (the learning) is only as good as the indexing assigned to the cases when they are stored. Another important learning issue is reindexing cases. When use of a case reveals some new knowledge, it should be reindexed to reflect that new knowledge (Kolodner 1993). CBR System Design, Maintenance, & Testing CBR System Design Some questions that must be considered when designing a CBR system are (Kolodner 1993): Is CBR the appropriate approach? Which portions of the system should use CBR, and what portions should use another approach? To what degree should the system be automated? 19


What level of support is needed a human reasoner if the system is not fully automated? System designers should keep in mind that the quality of the CBR system will depend on: the cases with which the system is "seeded" and its subsequent experiences; the system's ability to understand new situations in relation to past experiences; it ability perform/support adaptation; support for feedback; and its ability to appropriately add new experiences to the case base (Kolodner 1993} While there are several system design methodologies (HSTDEK, KEMRAS, KADS, and EX/Method} for building more traditional expert or knowledge-based systems, until recently, there was very little progress in this area for CBR systems. The only complete CBR methodology, CaseMethod, was developed by Kitano and Shimazu and was successfully applied in the development of SQUAD, a CBR system in use at NEC (Kitano 1996} Some tools, such as Design-MUSE, REPRO, and ASK, have started to emerge to aid in CBR system development. 20


Maintaining the Case Base Even if a quality effort is exerted to create a CBR system with an excellent initial case base, problems with the case base may arise. Several factors that may affect the quality of the case base over time are: novel situations that arise, but were not identified during initial testing; the system's purpose may evolve; normal experiences encountered in the domain may change; or, at the time of system design/fielding, the domain may have been new and not yet mature (Kolodner 1993). The key is to continue testing to ensure the right new cases are added to the case base, case coverage is complete, and that proper indexing exists. Testing Generally, testing specific to a CBR system involves testing to ensure the case base provides adequate coverage, identify inadequacies in the indexing scheme, and identify inadequacies in the contents of individual cases. Of course, this is an ongoing process necessary throughout the life of the system, although the need for testing may diminish over time depending on the maturity, stability, and complexity of the domain. System builders 21


should be systematic in collecting cases to "seed" a new CBR system. "Cases representing some set of common methods, solutions, and pitfalls in some area are enough to seed a case library. As those are used, it becomes clear what those cases don't cover and where the gaps are in the case library (Kolodner 1993). 22


CHAPTER 3 RULE-BASED EXPERT SYSTEMS, CBR, AND KNOWLEDGE MANAGEMENT Rule-Based Expert Systems An expert system (or knowledge-based system) is an interactive computer-based decision tool that uses both facts and heuristics to solve difficult decision problems in a specific area of interest, based on knowledge acquired from an expert (Badiru 1992). Such systems began to emerge in the mid-sixties and early seventies and have continued to achieve success, especially in the business, finance, education, manufacturing, telecommunications, legal, and medical domains. Rulebased expert systems are the most common type of expert system, but often expert systems are hybrid systems, incorporating more than one artificial intelligence technique. 23


Case-Based Reasoning In the late eighties and early nineties, case-based reasoning applications began to appear. Rather than replacing other expert system techniques, case-based reasoning complements existing techniques and tools such as databases, information retrieval, statistics, machine learning, neural networks, and rule-based systems. Casebased reasoning is proving especially beneficial when applied to problems in domains that are poorly understood, or that have many exceptions -domains where rule-based expert systems struggle. Take, for example, the commercial CBR advisor system, Clavier (Hennessy 1992). The Lockheed Corporation in Sunnyvale, California, creates low fault tolerance, aeronautical parts from sensitive and expensive composite materials. The parts are cured in a large oven, or autoclave, for which no theory or model of behavior exists. Successfully loading and curing a batch of various composite parts is a "black art" with each part having its own curing characteristics. Small parts are placed on tables inside the autoclave while larger parts, sometimes 20' by 50', are placed on the floor. All of the parts interact, altering the heating and cooling 24


characteristics of the autoclave. To make matters worse, faulty parts cannot be recycled for reuse of their expensive materials. A rule-based system was first attempted to help configure the layout of composite parts in the autoclave, but this failed due to poorly defined rules. Then development of Clavier, a CBR approach, began in late 1987 and has been in regular use since 1990. Clavier was originally fully automated, but was changed to give users control of adapting similar, past successful part layouts for current part layouts. Clavier was originally seeded with 20 cases and grew to several hundred cases. The system now retrieves a successful layout 90% of the time. Knowledge Management In today's corporate environment, employee turnover rates are growing higher and higher due to an aging work force, corporate restructuring and downsizing, a technical work force in high demand, and many other reasons. Every time an employee leaves, a part of the company's "corporate memory" leaves too. Industry's recognition of the value of corporate memory, and its treating corporate memory as an asset is called Knowledge Asset Management, or just Knowledge Management (Watson 25


1993). Knowledge Management includes reusing corporate memory (previous experiences and practices), with modifications, to satisfy present circumstances (Koulopoulos 1997) Carl Frappaolo of The Delphi Group, a leading enterprisewide knowledge management consulting firm, articulated the shift in corporate valuation from inventory to knowledge by pointing out that, "Microsoft-a company with a relatively minor physical plant and raw product inventory-is capitalized at $200 billion. By comparison, General Motors-a bastion of the Industrial Age, whose global facilities and inventory stand second to none-is capitalized at only $40 billion." Due to this recent shift of corporate emphasis toward valuing knowledge, knowledge management systems are starting to appear. An enterprise knowledge management system is a group of technologies and tools that facilitate the collection, organization, and transfer of knowledge among employees (Koulopoulos 1997). One such knowledge management tool is an automated warehouse of corporate knowledge called a corporate memory system or an Organizational Memory Information System (OMIS). According to Huhn and Abecker, the creation of corporate memory systems was motivated by the corporate success of database and hypertext systems. They also point out that Corporate memory systems store and supply relevant corporate information but, leave its 26


adaptation and/or reuse mostly to the user (Huhn 1998) This very closely parallels what CBR advisor systems do. CBR advisors are systems that leave the adaptation of cases retrieved from the case base entirely to the human user. Corporate Memory: A New Focus for CBR While knowledge management focuses on knowledge at the strategic level, expert systems focus on knowledge at the tactical level -within the scope of a specific problem domain. In both cases, the goal is to help nonexperts perform more like experts, and in fact, knowledge management systems sometimes include expert system subsystems. However, artificial intelligence could make more numerous and economical contributions by creating more CBR advisor systems designed specifically as corporate memories. The problem is that most of the CBR community's focus is on developing CBR systems that are fully automated while advisor systems receive much less attention. For example, in Kolodner's survey of CBR systems, 2/3 of the CBR systems are automated. Moreover, only 5/43 of the CBR systems are advisor systems (Kolodner 1993). 27


Take one more look at CLAVIER. While this system was originally designed to be fully automated, it was implemented as an interactive advising system. While Clavier does help Lockheed successfully create composite parts, Janet Kolodner points out, it also serves as "corporate memory" for Lockheed (Kolodner 1995) The key point being that if everyone quit on the same day, Lockheed could more quickly recover to continue to successfully make composite parts because, while the human experts left the company, their expertise remained. The design and creation of CBR systems specifically for knowledge management should be more common. In addition to under emphasis on CBR advisor systems in the CBR community, another reason that CBR advisors are not used more for corporate memory systems is that system designers are often using a rule-based approach for corporate memories. Case-based reasoning may be overlooked for creating corporate memory due to unfamiliarity with CBR, especially in the management information systems community. However, it is interesting that Kuhn's and Abecker's analysis of three case studies on rule-based corporate memory systems in Europe resulted in them concluding that in spite of considerable corporate interest in the proposed systems, "none of the three projects went beyond the prototype stadium, which makes it obvious that companies shun the risk and costs 28


of investing in novel technologies that have not yet found widespread application. This reservation is partly due to the unsatisfying results obtained from expert system technology with which all of the three companies had experimented in the past (Huhn 1998) ." The bottom line is that CBR is more suitable for creating corporate memory than the rule-based approach. Kuhn and Abekcer find their are five crucial requirements for corporate memory systems (Kuhn 1998) : 1. Collection and systematic organization of information from various sources (paper documents, electronic documents, databases, e-mail, CAD drawings, the personal notes and in the heads of individual employees, and so on). Collecting information from such a variety of sources poses problems regardless of the approach. However, CBR manipulates large chunks of data in the form of cases whereas rule-based systems represent knowledge in the form of rules. With CBR, information from any type format must be converted into cases. This is much easier than having to extract data from the same sources and convert the knowledge into rules. Since CBR doesn't use rules, but rather manipulates large chunks of knowledge in the form of cases, it has an advantage in manipulating a variety of data formats. The key for CBR with cases that contain multiple data formats becomes 29


proper indexing of the case data -regardless of the format in which it may be. 2. Integration into existing work environment (word processors, spreadsheets, CAD systems, simulators, etc.). Integration into existing work environments is perhaps an equal challenge for CBR or rule-based systems. However, CBR tools, such as the one from Esteem Software, support DDE access application embedding and multimedia data types such as sound, photographs, and video. 3. Minimization of up-front knowledge engineering. Case-based reasoning typically avoids the knowledge elicitation bottle neck associated with rule-based systems. While acquiring case data may sometimes be a challenge depending on what amount and format of data exists, the process of gathering knowledge (cases) for a CBR system is significantly easier than that required for the elicitation of knowledge for the creation of rules for rule-based systems. 4. Active presentation of relevant information (actively remind workers of helpful information and be a competent partner for cooperative problem solving). This, almost by definition, is what a CBR advisor system does. Furthermore, CBR systems provide precedence in guiding users while rule-based systems provide a series of rules that fired for justification. In an interactive advising system such as a corporate memory, guidance to the user 30


in the form of case precedence is more desirable than being provided a sequence of rules that fired. 5. Exploiting user feedback for maintenance and evolution (minimum maintenance and ability to deal with incomplete, potentially incorrect, and frequently changing information) Again, CBR has the advantage over rule-based systems in meeting these requirements. While frequently changing information has a negative impact on both approaches, adding and deleting cases from a case base to keep abreast of change is easier than constantly modifying a complex rule system. Rule-based systems do not handle incomplete, or fuzzy, information well at all. On the other hand, a CBR system that does not perform automated adaptation, can better handle fuzzy data. Once again, rule-based systems are inferior to CBR systems at handling incorrect data. Case-based systems with a healthy case-base will be able to provide alternatives to maneuver around incorrect data where a rule-based system will typically fail and perhaps point out the problem if such a problem was anticipated. CBR Knowledge Management Systems Currently, only a handful of companies such as Texas Instruments, Apple Computers, and NEC have created CBR 31


systems specifically to manage their employee's experience and knowledge. For example, SQUAD is NEC's CBR software quality control system to manage a corporatewide case-base of more than 20,000 cases that provide solutions or just warn of pitfalls. The case base grows by about 3,000 cases each year (Kitano 1992). While businesses would probably like to retain most all of their employee's experience and knowledge if they could, this is not practical simply based on the magnitude of the problem. While SQUAD is a large-scale, corporate-wide, CBR system, it is only for software quality control, not multiple domains. Thus, the issue of retaining corporate knowledge becomes defining when is it worth while to invest in a case-based reasoning knowledge management system. Criteria for Applying CBR to Knowledge Management I propose CBR corporate memory systems are applicable in all cases where: the domain is not well defined or numerous exceptions exist; versus computers, humans can still more efficiently adapt past experiences to solve problems; there are not enough experts; and there is some experience within the domain. 32


Domain Type If the problem domain is not well defined or numerous exceptions exist, then traditional expert system approaches are not ideal. In particular, rule-based systems require the solution to be known and fully understood for the creation of the system's rules. There are many domains, such as the creation of composite parts in an autoclave, where this is not practical, or even possible. Case-based reasoning can succeed in such domains as long as the rate of change is not too rapid regarding the principles that hold true within the domain. If the fundamental truths associated with the problem domain change too rapidly, then maintaining an adequate case base becomes a problem. User Adaptation A case-based reasoning or hybrid expert system can be created if the domain is defined well enough to allow automation of case adaptation. However, a problem domain should not be abandoned if, versus computers, humans can 33


still more efficiently adapt past experiences to solve problems. In these cases where a traditional expert system or an automated CBR system is not well suited, a case-based reasoning advisor system can be developed to improve employee performance simply by serving as a corporate memory for non-experts. Often, simply knowing what did and did not work in the past allows humans to reuse this knowledge to improve performance. Domain Experts If there is a problem domain where there are not enough experts to meet needs and the domain is not conducive for traditional expert systems or automated CBR systems, then a CBR corporate memory system may be created to help non-experts improve performance. If there are an adequate number of experts, a corporate memory may still prove beneficial depending on the domain's broadness and complexity. Domain Experience There must be some experience within the domain for a case-based reasoning system of any type to be "seeded" 34


with a minimum number of cases. Depending on the domain, a CBR system may require only a handful or several hundreds of cases at the system's inception. The key is that enough experience must exist for cases to be added to the case base to cover solutions or approaches to standard problems, and any novelties encountered in the domain. Through system operation and feed back, the case base can then grow and learn as new cases are gathered about problems that may become standard and as new or overlooked novelties are encountered. 35


CHAPTER 4 BRAC ADVISOR: AN EXAMPLE CBR CORPORATE MEMORY SYSTEM The Problem Retaining and disseminating the military's BRAC knowledge is the most critical of all BRAC problem areas. The BRAC process is not continuous, but rather periodic. Several rounds of BRAC have been executed and more rounds are projected, but years pass between occurrences. As a result, experts are lost, moving on to fill other duties and even different careers outside of the shrinking realm of the Department of Defense. Further complicating this loss of experts is the standard policy for military personnel not to serve in repetitive assignments, but to seek jobs of increasing responsibility in a variety of areas. So, even if the BRAC experts are still with the military, they are often not involved in subsequent rounds of closure. Not only is expertise lost, it is also not effectively shared. Formalized collecting of 36


knowledge or lessons learned is not done at the DoD or even DA level. Some major commands within the DA collect and distribute such documentation, but on the whole, there is no standard effort, scope, format, or reuse of past experience. This is particularly unfortunate given that the staff at BRAC sites are told that in order to succeed, they must (BRIM 1995) : Be flexible. Do not be wedded to a particular approach to a problem. Creativity, within the applicable laws and regulations, is essential to successful base closure and reuse. Cut ured tape". The Department of Defense is seeking to eliminate as much red tape as possible. By itself, each additional procedure may make sense, but the accumulation of them over time grinds matters to a halt. Cutting across bureaucratic lines to slash red tape is essential to getting the job done and additional levels of bureaucracy can lead to higher costs and wasted taxpayer money. Be innovative. Do not be hamstrung by past practices. This is a new process, and decisions should be new and different. Exercise common sense. Solutions should be sitespecific. There will rarely be cookie-cutter solutions that apply to all cases. Such creative solutions and possible short cuts should be shared for recalling among current BRAC sites and preserved for future sites. 37


Another reason this knowledge should be captured and shared is the local military base commander is the primary executive in charge of leadingand managing the BRAC Process. Therefore, it is almost ensured that the key decision maker at a BRAC site will be a novice in terms of BRAC -as will be the base commander's staff. To help alleviate this situation, the commander of a BRAC site is authorized a Base Transition Coordinator, someone formally trained in the closure process, but who comes without any guarantee of having adequate, previous BRAC experience. Matters are further compounded by the fact that the community, with whom the base commander must execute the closure or realignment in coordination with, will almost certainly be novice to the BRAC process as well. The DoD'sBase Reuse Implementation Manual warns novice BRAC site commanders and their staffs that, "[t]he base reuse process is affected by a myriad of Federal real property and environmental laws and regulations, along with volumes of implementing guidance {BRIM 1998) ." Added to this is strong encouragement to be creative and to find short cuts that make sense within the limits of the law. Certainly, the above is substantial justification for maximum distribution and thoughtful preservation of BRAC knowledge. A proposed CBR solution to address this 38


particular problem follows. Additional BRAC problems areas that can benefit from computer science are identified in appendix D. A Proposed CBR Solution The BRAC process is decentralized in that commanders at individual BRAC sites are the military's primary decision maker on planning and executing how each base will close. This is necessary as each base is unique. However, base commanders and their staffs are virtually all novices to the BRAC process. Commanders receive assistance from the DA and DoD BRAC Offices, but experts in these organizations are limited as well due to workload and personnel turnover. To further complicate matters for novice base commanders and their staffs, the BRAC process is governed by numerous laws, and volumes of guidance on interpretation of the law is provided. Moreover, military commanders are encouraged to be innovative and creative within the BRAC laws and told to cut "red tape" and cross bureaucratic lines whenever possible (BRIM 1998) A case-based reasoning system should be created to serve as corporate memory and to fill the void of experts and expert systems in the BRAC process, helping base 39


commanders as well as DA and DoD BRAC offices. I propose a case based reasoning system be created that I call BRAC Advisor. Unlike the few existing human BRAC experts, BRAC Advisor will be permanent. It will not quit, change jobs, retire, or die. It will be there whenever Congress mandates new rounds of BRAC. It will also never stop learning as long as it is actively used in the BRAC process, nor will it unintentionally forget what it knows of past rounds of BRAC. The BRAC Advisor can be in many places at once, serving as an expert to (and learning from) all BRAC sites simultaneously, thereby allowing instantaneous sharing of innovations, creativity, successful cutting of red tape, and crossings of bureaucratic lines. Criteria for Applying a CBR Corporate Memory I propose CBR corporate memory systems are applicable in all cases where: the domain is not well defined or numerous exceptions exist; versus computers, humans can still more efficiently adapt past experiences to solve problems; there are not enough experts; and there is some experience within the domain. The problem domain of military base closure operations that I have 40


chosen for demonstrating the benefits of a case-based reasoning corporate memory meet these applicability criteria. Domain Type The domain of BASE closure operations is not well defined. While the domain is governed by volumes of laws and regulations, the fact that those carrying out base closure are encouraged to be creative and to cut bureaucratic "red tape" makes the domain a little fuzzy. Furthermore, each base closure site is very unique. The closing base may by a Naval ship yard, and Army ammunition or repair depot, or an Air Force base supporting a wing of bomber aircraft or a field of nuclear missile silos. Adding more uniqueness to the closure of a base site is the civilian community associated with each closing base. Some communities may be affluent, while others may be impoverished areas. Waivers and exceptions to policies may be granted to aid a community that suffer tremendously due to the loss of government jobs that will result from base closure. All these exceptions to the rules make case-based reasoning an ideal approach for base closure. 41


User Adaptation Due to the incredible uniqueness that each base closure presents, there is no "cookie cutter" solution to how a base should be closed. In a problem domain where there are so many exceptions, automating the adaptation of past experiences is not practical. Automating adaptation generally requires the use of rules, which due to the very nature of the problem domain are not suited for use. Moreover, when special considerations such as creativity, historical issues, and compassion must be weighed when choosing an adaptation path, human adaptation is currently better than automation. Domain Experts Again, there are few BRAC experts due to military assignment policy, the period of time between rounds of BRAC being years, and the ongoing reductions in the Federal civilian work force. 42


Domain Experience There is plenty of experience within the domain as there have been four rounds of base closure. The question then becomes uis this experience been documented in some way so it can be captured and reused?". The answer is yes. While there is no formal BRAC central collection point for BRAC experience, individual closures sites almost always create some sort of paper or electronic document containing ulessons learned." In addition, there are some experts in the field who remain and can share their knowledge. General Advantages of Using CBR for This Problem Watson identifies several characteristics of domains where CBR is most applicable (Watson 1997) Some of these characteristics that apply to BRAC are: Domain experts routinely compare a current problem to past cases. Experts already adapt cases to solve new problems 43


Cases are available in bibliographic sources and in experts' memories, and can be recorded as new solutions are generated. There are means in the domain to assign an outcome to a case, explain it, and deem it a success or a failure. Cases can be generalized to some extent. Features that make them relevant can be abstracted. Cases retain currency for a relatively long time. Cases are used in training professionals in the domain. Base Realignment and closure involves complex, structured data that changes slowly with time. Justification is extremely important as closing installation commanders make multi-million dollar decisions that by nature are extremely sensitive and political. Case data is available, complex adaptation is required, but automated adaptation is not, and an exact or optimum answer is not required. Rule-based, and even ther expert system approaches, simply do not exceed or even match the benefits of using CBR in such a domain. Case-based reasoning's advantages in domains that are not understood or that have many 44


exceptions have already been discussed. But take, for example, justification. Rule-based and machine learning approaches reference rules for justification which is not as understandable as concrete examples or precedence. Other approaches like neural networks cannot really do any justification. When it comes to complex and fuzzy data, CBR outperforms database and information retrieval approaches. Case based reasoning avoids the knowledge elicitation bottle neck. Novices can simply enter cases from current and past BRAC sites into the knowledge base, allowing the system to learn. BRAC Advisor System Design Choosing a CBR Tool The BRAC Advisor will simply serve as an advisor system. Since no automated adaptation occurs, using one of the many CBR tools that have become available makes sense. Support for multimedia may also be a factor in deciding which tool to use. Any CBR tool selected may require some programming to convert any existing databases with useable data for cases. Also, programming may be needed to make the system web based. Support for 45


these issues is also important in considering a tool. My first choice for a CBR tool is ESTEEM from Esteem Software. It runs on a PC Windows platform, it includes a simple, form based GUI builder, supports multimedia, and only costs about $500. Other nice features are that its representation scheme allows nested cases, it can perform some adaptation through functions and rules. It uses nearest neighbor with inductive weight generation for retrieval. My second choice is CaseAdvisor from Sententia Software. It doesn't support adaptation, but that is not needed. It uses Netscape for a user interface and has a WebServer module. It runs on both PC Windows and MVS platforms and costs about $1,000. It uses flat records for representation and allows weighting. It uses nearest neighbor retrieval. CASPIAN Due to cost considerations, what I actually used is CASPIAN, a public domain CBR tool from the University of Aberystwyth in Wales. It is written in C and can run on MS-DOS or Macintosh. It uses a simple command line interface, but can be integrated with a GUI. It performs simple nearest neighbor retrieval and does allow adaptation through rules in an ASCII file. 46


Case design in CASPIAN is limited to the following format. The examples are from the CHEF CBR system which is an automated CBR system that can create new receipts from its case base of receipts. case instance mange_tout is ingredients = [ green_beans ] ; cook_method = stir_fry; taste = savory; vegetable = green_beans; solution is recipe = [ cook method the ingredients ] ; end; A more advanced case that includes case repair follows: case instance chicken_green_beans_stir_fry is cook_method = stir_fry; ingredients = [chicken green beans] ; taste = sweet_n_sour; meat = chicken; vegetable = green_beans; solution is meat_preparation = bone; vegetable_preparation = shell; recipe= [[meat_preparation the meat] [vegetable_preparation the vegetable] [cook_method the meat and the' vegetable together using a wok']]; 47


local repair rule definition is repair rule reselect_l is when then end; meat is undefined pr(['Abandoning case because meat is undefined']); reselect; repair rule reselect 2 is when then end; vegetable is undefined pr(['Abandoning case because vegetable is undefined']); reselect; repair rule prepare_broccoli is when vegetable is broccoli then change vegetable_preparation to chop; change hd(tl(tl(tl(tl(tl(hd(tl(tl(recipe))))))))) end; to separately; pr(['Warning: The' meat 'makes the' vegetable 'soggy']); pr(['when they are stir-fried together, so cook']); pr(['the' meat 'and the' vegetable 'separately']); repair rule prepare_legume is when vegetable is legume 48


then end; end; change vegetable_preparation to shell; CASPIAN & BRAC Advisor The cases in BRAC Advisor contain knowledge related to disposing of real and personal military property. Real property includes buildings, roads, grounds and other fixed assets. Personal property includes non-fixed or non-permanent items of all sizes (from trains to paper clips and staples) BRAC Advisor case problem definitions contain the following fields: category of reuse (the policy under which the conveyance of military property for reuse by a non-military agency is authorized) category of property (real and/or personal property) type of property (such as medical, police, or office furnishings) The BRAC Advisor case-base and control data for CASPIAN is included in appendix A. Figure 3-1 shows what categories of property are represented in the case base. 49


Each path from the top of the figure to the bottom of the figure represents a case. Paths that dead-end before reaching one of the types of property at the bottom of the figure are not cases in the case base. For example, one can trace down along a path from any one of the categories of conveyance along the top of Figure 3-1 down to one of the types of property to determine the problem field values of a case that is in the BRAC Advisor case base. So, the case base contains a case with the following problem description field values: Category_of_Reuse = Public_Airport_Conveyance Category_of_Property = Real Type_of_Property = Aviation However, the case base does not contain a case with the following problem field values: Category_of_Reuse = Economic_Development_Conveyance Category_of_Property = Real_and_Personal Type_of_Property = Medical Figure 3-2 shows what categories of reuse are in the case base. These figures may be useful to help see what so


cases were retrieved and why when looking at the sample run of BRAC Advisor in appendix B. 51


lJ1 tv BRAC Advisor Cases (Categories of Property 11) Long Term Lease in Furtherance of Conveyance Medical Entertainment & Recreation Housing


U1 w BRAC Advisor Cases (Categories of Reuse [I ) Lease in Furtherance o!C7eyance Real& Sponsored hi& Personal Personal Real& Personal Real&


CHAPTER 5 CONCLUSIONS Case-Based Reasoning Case-based reasoning is certainly on the rise as a new, viable approach within artificial intelligence. More and more CBR successes are occurring, not only in the academic world, but in the commercial world as well. Major areas of research that will further improve CBR's usefulness and are the creation of system design methodologies and a robust pool of CBR tools that adequately support case collection, indexing, and maybe even better adaptation. Scaling up is another critical area that needs solutions. Problems arise with system retrieval capabilities when moving from a case base of 100 cases to 1,000, or even tens-of-thousands of cases. In addition to knowledge management, knowledge navigation is another area of research that I found interesting and for which CBR can provide great benefit. Knowledge navigation uses CBR to characterize and guide 54


information search which will be very useful in applications such as searching/browsing on-line repositories of information. Information can be gathered by referring to a specific example -similar to when a person sees a car at a dealership and tells the salesperson, "something like that, but a little sportier (Leake 1996) ." CBR for Knowledge Management There is some indication that CBR will gradually become used more for knowledge management. Again, there are a handful of CBR systems fielded that were designed to support knowledge management. And there are researchers, like Kitano and Shimazu who developed NEC's SQUAD system, who are drawing attention to the need to divert some focus from adaptation and place it on knowledge management. They point out that CBR systems have traditionally been used as problem solvers and acknowledge the importance of this approach, but "argue that there is an area where CBR techniques can exhibit a far larger economic impact ... [in addressing] the practical and emergent need of business management to facilitate the sharing of corporate-wide knowledge (Kitano 1996) ." Of great benefit to CBR knowledge 55


management, and CBR in general, will be advances in CBR's ability to support human like cross-domain remindings of similarities between experiences. Lastly, the BRAC Advisor prototype should lead to the development of a real-world system to capture and disseminate the military's "corporate knowledge" regarding BRAC. The system should be web based and managed at the Department of Army or Department of Defense BRAC Office levels. 56




-BRAC95.txt is a case file containing sample DA BRAC cases -where exceptionally creative approaches were applied, critical -problems were encountered, or important precedents were -established. The purpose of this case file is to demonstrate the -advantages of using CBR techniques simply for knowledge asset -management. -NOTE: Abstractions in the Modification Section must be updated when -case definition field enumerations are updated. introduction is 1 I I I I I I I I I 1BRAC Advisor Demo -I I I I I I I I I I 1 end; -The following case definition defines the fields used in the -problem section of a case. case definition is field category_of_property type is (real, personal, real_and_personal, any_property_category) weight is 0; field category_of_reuse type is (any_type_lease, interim_lease, long_term_lease, lease_in_furtherance_of_conveyance, public_airport_conveyance, any_type_public_benefit_conveyance, sponsored_public_benefit_conveyance, approved_public_benefit_conveyance, homeless_assistance_conveyance, economic_development_conveyance, any_type_sale, direct_sale, public_sale, any_category_of_reuse) weight is 0; field type_of_property type is (any_type, administrative, aviation, auditorium_or_theater, automation, communication, entertainment_and_recreation, fire_fighting, housing, industrial_or_manufacturing, maintenance, medical, multiple_types, police_or_security, religious, roads_and_grounds, 58


end; transportation, undeveloped_land, utilities, warehousing) weight is 0; -There must be at least one field which is defined to be -an index field. Index fields must be enumeration type. index definition is end; index on category_of_reuse; index on category_of_property; index on type_of_property; -The following section defines a symbol hierarchy, which specifies -which symbols are similar. In addition, abstract symbols match -against their children when repair rules are used and during case -matching. This section may also be used to define ranges for -numeric fields to define when numbers are similar. modification definition is abstraction any_category_of_reuse is (any_type_lease, interim_lease, long_term_lease, lease_in_furtherance_of_conveyance, public_airport_conveyance, any_type_public_benefit_conveyance, sponsored_public_benefit_conveyance, approved_public_benefit_conveyance, homeless_assistance_conveyance, economic_development_conveyance, any_type_sale, direct_sale, public_sale); abstraction any_type_lease is (interim_lease, long_term_lease, lease_in_furtherance_of_conveyance) ; abstraction any_type_public_benefit_conveyance is (sponsored_public_benefit_conveyance, approved_public_benefit_conveyance) ; abstraction any_type_sale is (direct_sale, public_sale); 59


end; abstraction any_property_category is (real, personal, real_and_personal); abstraction any_type is (administrative, aviation, auditorium_or_theater, automation, communication, entertainment_and_recreation, fire_fighting, housing, industrial_or_manufacturing, maintenance, medical, multiple_types, police_or_security, religious, roads_and_grounds, transportation, undeveloped_land, utilities, warehousing) ; -THE FOLLOWING ARE CASES: case instance CAIRWIN_PBC_PM0_1 is category_of_reuse = sponsored_public_benefit_conveyance; category_of_property = personal; type_of_property = police_or_security; solution is end; base = ['Fort Irwin, CA']; round= ['1993']; source = ['BRAC93-CAIRWI']; summary ['The PMO compound at Fort Irwin was conveyed']; contd 1 ['through the US Dept of Justice to the']; contd 2 ['California Dept of Justice as a new prison.']; contd_3 ['An attempt was made to convey early the']; contd 4 ['personal property associated with the PMO']; contd_S ['real property compound by "washing" the']; contd 6 ['personal property through DRMO and theCA']; contd 7 ['state surplus agency. DRMO was unable to']; contd 8 ['do this as GSA would not grant a waiver.']; 60


case instance COFITZ_PBC_500_1 is category_of_reuse = sponsored_public_benefit_conveyance; category_of_property = real_and_personal; type_of_property = medical; solution is end; base = ['Fitzsimons Army Medical Center, CO']; round = ['1995']; source= ['BRAC95-COFITZ']; summary ['The main bldg of a 1,000 bed capacity']; contd 1 ['Army MEDCEN was conveyed through the']; contd 2 ['US Dept of Education to the University']; contd 3 ['of Colorado Health Science Center, a']; contd_4 ['state operated medical school. Larger']; contd 5 ['plan is for the med school to attract']; contd_6 ['companies to bio-science research park.']; contd 7 ['Personal property was conveyed separately'); contd 8 ['and early through a US Dept of Edu waiver.']; case instance COFITZ_UPARCEL_PBC_1 is category_of_reuse = sponsored_public_benefit_conveyance; category_of_property = personal; type_of_property = multiple_types; solution is end; base = ['Fitzsimons Army Medical Center, CO']; round = ['1995']; source = ['BRAC95-COFITZ']; summary contd 1 contd 2 contd 3 contd 4 ['Exception to policy granted from David']; ['Hacola of US Dept of Education for conveyance']; ['of personal property to the University of']; ['Colorado Health Science Center before the']; ['associated real property was conveyed.']; 61


case instance COFITZ_GOLF_EDC_1 is category_of_reuse = economic_development_conveyance; category_of_property = real_and_personal; type_of_property = entertainment_and_recreation; solution is end; base= ['Fitzsimons Army Medical Center, CO']; round= ['1995']; source = ['BRAC95-COFITZ']; summary contd 1 contd 2 contd 3 contd 4 contd 5 contd 6 contd 7 contd 8 contd 9 contd 10= contd 11= ['Golf course was operated by the City of']; ['Aurora as a public course until later']; ['redevelopment as a bio-science research park.']; ['Key issues revolved around the sprinkler'); ['system installed below ground and who would'); ['maintain the grounds during winter months'); ['after Army staff left and before LRA could'); ['operate to make money. Difficulties were'); ['encountered in determining whether the']; ['sprinkler was NAF/AF and its value. LRA']; ['argued zero cost based on disposal cost'); ['to the Army if not given free to LRA. '); case instance COFITZ_249_EDC_1 is category_of_reuse = economic_development_conveyance; category_of_property = real_and_personal; type_of_property = warehousing; solution is end; base = ['Fitzsimons Army Medical Center, CO']; round= ['1995']; source = ['BRAC95-COFITZ']; summary contd 1 contd 2 contd 3 contd 4 contd 5 ['The warehouse was used by the base commissary']; ['which continued to use the warehouse when the'); ['Redevelopment Authority assumed control. The'); ['commissary leased the warehouse from the LRA'); ['until the LRA needed the warehouse for long-'); ['term redevelopment.']; 62


case instance COFITZ_PBC_531_1 is category_of_reuse = approved_public_benefit_conveyance; category_of_property = real; type_of_property = fire_fighting; solution is end; base = ['Fitzsimons Army Medical Center, CO']; round= ['1995'); source = ['BRAC95-COFITZ'); summary contd_1 contd 2 ['Facility was conveyed to the City of Aurora'); ['for use as an active fire station. Personal'); ['property was obsolete and turned in to DRMO.'); case instance COFITZ LIFC 500 1 is ---category_of_reuse = lease_in_furtherance_of_conveyance; category_of_property = real_and_personal; type_of_property = multiple_types; solution is end; base = ['Fitzsimons Army Medical Center, CO'); round= ['1995'); source = ['BRAC95-COFITZ']; summary ['Former Army Medical Center main hospital bldg'); contd 1 ['was put under LIFC to LRA. This large bldg'); contd_2 ['later became occupied by several reuse'); contd_3 ['activities which were executed as sub-leases'); contd 4 ['to the main, original lease. This caused'); contd_S ['problems/restrictions later when attempting'); contd_6 ['to modify/eliminate the main lease. Changes']; contd_7 ['were not made because of overhead in changes'); contd 8 ['to all sub-leases. Keep leases modular.'); 63


case instance COLOWE_PBC_AIRFIELD_l is category_of_reuse = public_airport_conveyance; category_of_property = real; type_of_property = aviation; solution is end; base = ['Lowery']; round= ['1993']; source = ['BRAC93-COLOWE']; summary ['The Lowery AFB airstrip, several hangers,']; contd_l ['and the terminal were conveyed to the City']; contd 2 ['of Aurora for their police department driver'); contd 3 ['training and vehicle maintenance.']; case instance COFITZ_HAC_300_1 is category_of_reuse = homeless_assistance_conveyance; category_of_property = real_and_personal; type_of_property = housing; solution is end; base= ['Fitzsimons Army Medical Center, CO']; round= ['1995']; source= ['BRAC95-COFITZ']; summary contd 1 contd_2 contd 3 contd 4 contd 5 contd_6 contd 7 contd 8 contd 9 ['Army barracks was closed down as military']; ['staff reduced. Barracks with furnishings']; ['were eventually conveyed to homeless']; ['assistance program. However, the USAF was']; ['allowed to use the barracks during interim.']; ['This resulted in problems when USAF failed']; ['to meet timeline to move out of facilities'); ['and also due to wear and tear on furnishings'); ['that has previously been inspected by the'); ['homeless assistance program chief.']; 64


case instance COFITZ_DIRSALE_700_1 is category_of_reuse = direct_sale; category_of_property = real; type_of_property = housing; solution is end; base= ['Fitzsimons Army Medical Center, CO']; round= ['1995']; source= ['BRAC95-COFITZ']; summary contd 1 contd 2 contd 3 contd 4 ['The government quarters were sold directly']; ['to the City of Aurora which plans to rent']; ['the quarters primarily to students at the']; ['Univ of CO Health Science Center which moved']; ['onto the base as part of the redevelopment.']; case instance COFITZ_PUBSALE_CLUB_1 is category_of_reuse = public_sale; category_of_property = real_and_personal; type_of_property = entertainment_and_recreation; solution is end; base = ['Fitzsimons Army Medical Center, CO']; round = ['1995']; source = ['BRAC95-COFITZ']; summary ['A public sale/auction was held to sell the']; contd 1 ['community club that had recently been built']; contd 2 ['on the base. Disagreements arose between']; contd 3 ['the Army and the LRA over conflicting values']; contd 4 ['for club worth submitted by each agency']; contd 5 ['appraiser. Problems could have']; contd 6 ['been reduced if a standard inventory of what']; contd 7 ['property was for sale was used by both']; contd 8 ['appraisers. A key problem was determining']; contd 9 ['what was real and what was personal property.']; 65




SAMPLE RUNS OF BRAC ADVISOR Before beginning with the samples, the reader may want to look at appendix C for a BRAC process overview. For these sample runs, assume that the user of BRAC Advisor is a commander at a closing military installation, Fort Resort, located in the City of Closureville. The City of Closureville has been negotiating for some time to take ownership of several of the nicer buildings on Fort Resort. The commander has a large administrative building on her installation that no agency has expressed interest in due to its poor condition, but the building contains nice office furnishings. The City of Closureville has said they would like to have the nice furniture in the rundown building, but they will not take the building. The city is getting other buildings and furnishings from Fort Resort for free through a public benefit conveyance sponsored by the us Department of Labor. The commander wants to know if there is some way she can give just the furniture out of the rundown building to the city for free under some type of public benefit conveyance. The commander starts BRAC Advisor. The commander wants to search for matching cases that only involve personal property, which is how the military classifies the furniture. 67


'"""'"'"""'""'"""'""""""'""""""'""""""'"""'"""'"""""""'"""""""'""'""'"""""''''''''''''"''''":::.:.:.:: . . ,. . . ' -. .. -) 1111111111-BRAC Aduisor Demo -1111111111 Search for matching case Search specifying indexes separately Turn list expansion ON/OFF Quit choice (1,2,3 or 4): 1 The commander chooses to search the case base (1) :. C:\TEMPI.CASPrim:ExE . . . . : =.. ![ .i -. "" . . . . I B !select ualue for category_of_property i I I 1) real 2) personal 3) real_and_personal 4) any_property_category choice (1 to 4): 2 The commander is looking for cases that exclusively involve the transfer of personal property (2) 68


ualue for category_of_reuse 1) any_type_lease 2) interirn lease 3) long_terrn_lease 4) lease_in_furtherance_of_conueyance 5) public_airport_conueyance 6) any_type_public_benefit_conueyance 7) sponsored_public_benefit_conueyance 8) approued_public_benefit_conueyance 9) horneless_assistance_conueyance 10) econornic_deueloprnent_conueyance 11) any_type_sale 12) direct sale 13) public=sale 14) any_category_of_reuse choice (1 to 14): 6_ The transfer must be free to the City of Closureville, so the commander selects (6) any type of public benefit conveyance. 69


..::::.::::::::.:::::.:::.::::::::::::::::::::::.::::::::::::::::::::::::::::::::::: Select value for type_of_property 1) any_type 2) administrative 3) aviation 4) auditorium or theater 5) automation-6) communication 7) entertainment and recreation 8) fire_fighting-9) housing 10) industrial_or_manufacturing 11) maintenance 12) medical 13) multiple_types 14) police_or_security 15) religious 16) roads_and_grounds 17) transportation 18) undeveloped_land 19) utilities 20) warehousing commander is looking to convey office furnishing, but loes not select administrative type (2) property. Rather, :he starts with a broader category (1) to increase her of finding some creative way that has been used .n the past to transfer only personal property under some :ype of public benefit conveyance. Note: CBR does allow attributes, providing more powerful retrieval :han databases. BRAC Advisor returns the following two cases it 1redicts will be usefully relevant for the commander. 70


o::::::::::::o:.:.::::::.:.:.::,::.:,:.::;::""""""'"'"' ase instance CAIRWIH_PBC_PM0_1 is category_of_reuse = sponsored_public_benefit_conueyance; = type_of_property = police_or_security; olution is base= [ 'Fort Irwin, CA' ]; round= [ '1993' ]; source= [ 'BRAC93-CAIRWI' ]; summary [ 'The PMO compound at Fort Irwin was conueyed' contd_1 [ 'through the US Dept of Justice to the' ); contd_2 [ 'California Dept of Justice as a new prison.' contd_3 [ 'An attempt was made to conuey early the' ]; contd_4 [ property associated with the PMO' ]; contd_5 = [ property compound by "washing" the' ]; contd_6 [ 'personal property through DRMO and the CA' ); contd_7 [ 'state surplus agency. DRMO was unable to' ); contd_B [ 'do this as GSA would not grant a waiuer.' ); ... -: '. ,::.: .. . .. :: .-: : ,;:;.: .. : >""::--: ,, ....... ,. : .. : ' ... . ....... .. .:. ., . ... .. :_:.::.-:.: .: .. 71


instance COFITZ_UPARCEL_PBC_1 is category_of_reuse = sponsored_public_benefit_conueyance; category_of_property = personal; type_of_property = multiple_types; :; olution is .: base= [ 'Fitzsimons Army Medical Center. CO' ]; round= [ '1995' ]; source= [ 'BRAC95-COFITZ' ]; summary [ 'Exception to policy granted from Dauid' ]; contd_1 [ 'Hacola of US Dept of Education for conueyance' contd_2 = [ 'of personal property to the Uniuersity of' ]; contd_3 [ 'Colorado Health Science Center before the' ]; contd_4 = [ 'associated real property was conueyed.' ]; Select the displayed case Co Back to preuious case look at Next case Neither of the two cases provided by BRAC Advisor directly match the commander's problem. However, she now has learned several valuable things. Personal property may be transferred early with a waiver from the sponsoring Federal agency. The other thing the commander learned is what does not work. She can go tell her subordinate who had started searching for creative solutions to convey the furniture not to waste his time perusing a "wash" transaction through DRMO because the 72


agency, GSA, will .not grant exceptions to policy. Unfortunately, the commander still doesn't know if personal property can be transferred separate from the real property it is in or associated with. On the bright side, the knowledge about whether personal property can be transferred separately from its associated real property will be added to the case base as soon as this problem is resolved. One more example. The commander has a golf course on her installation. The City of Closureville wants the course, but cannot decide exactly what they will do with it. Some want to develop the land commercially, and others want to run the golf course as a city golf course. Recently, she heard the city was talking about even more options. So the commander turns to BRAC Advisor again. This time, the commander just wants to see what the issues have been at any other BRAC sites that may have had golf courses so she can perhaps steer the city's decision or at least be prepared for some of the problems that may be common to disposing of a golf course. 73


!1 !select ualue for category_ of _jlroperty :lil 1ip1 !! I The commander chooses to search for a case again, but this time she leaves the category of property open (4), since the golf course has all types of property and she wants to learn of any issues that may arise 74


.......................................................................................................................................................................................... for category_of_reuse r..-.'_,_',_l.i_:._: 1) t 1 -any_ ype_ ease '\N !iii C:\lEMP\CASPIAt.iXE .. . . 2) interim_lease 3) long_term_lease 4) lease_in_furtherance_of_conueyance 5) public_airport_conueyance 6) any_type_public_benefit_conueyance :: 7) sponsored_public_benefit_conueyance !i 8) approued_public_benefit_conueyance 9) home less_ assistance_ conueyance !I 12) direct sale I IIEnter choice (1 to 14): 14 Again, the commander doesn't know what the city will do with the course, so she doesn't know how they will ask that it be conveyed to them. She chooses (14). 75


. "'""1 l !!select ualue for type_ of _property w! 3) auiation ==''='' !! 4) auditoriu111 or theater ;)81 I and recreation 8) fire_fighting-9) housing 10) industrial_or_ITianufacturing 11) 111aintenance 12) 111edical 13) 111ultiple_types 14) police_or_security 15) religious 16) roads_and_grounds 17) transportation 18) undeueloped_land 19) utilities 20) warehousing !!Enter choice (1 to 20): 7 i:::=::::lj The commander chooses entertainment and recreation (7) since she believes this is what cases involving a golf course will be indexed under. BRAC Advisor returns the cases below to the commander. The commander learns from the first case that if the golf course cannot be conveyed for free under some public benefit, but most be sold, then appraisers and negotiations will be necessary. The commander was aware of this, but the issue of agreeing to a standard inventory before valuation by the appraisers is something she had not thought of. Also, she is concerned about the 76


problems other BRAC sites had in the past where parties were not able to distinguish or agree upon the different types of property. The commander is really lucky. There is a golf course related case. From this she learns of maintenance issues and sees again that determining the types of property is an issue. This is a matter she can address now while the city is deciding what their plan will be. She will share with the city at her next meeting the idea of interim use of the course if the city decides to develop the land for commercial use. 77


''"::""::::::<:::::::::::::::::,::::::::::::::::::::::::::::::::::::::::::::::::::::::::;:::::::::::::::::::::::::::::::::::::::::,,:::::::::::::::::::: .......................................................... instance COFITZ_PUBSALE_CLUB_1 is category_of_reuse = public_sale; category_of_property = real_and_personal; type_of_property = entertainment_and_recreation; is base = [ 'Fitzsimons Army Medical Center, CO' ); round = [ '1995' ); source = [ 'BRAC95-COFITZ' ); summary ( 'A public sale/auction was held to sell the' ); contd_1 ( 'community club that nan ); contd_2 ( on the base. ) ; contd_3 [ the Army and theo .. I::RA .. oueot .. ctmfl:icting ualues ) ; contd 4 [ for club ageney ] ; contd_5 [ 'appraiser. &&Hld have' ]; contd_6 ( been reduced if a -+taRdar-d-i-Au.e-ntury of what' ) ; contd_7 [ 'property was -E.or.: ... sale ... bfas ... used ... by both' ).; contd_B [ 'appraisers. A key problem was determining' ); contd 9 [ what was reaL.and ... l!lllat ... was. ... p.e.r.sonal property. ) 78


ffi .. .. .... : . . . ... .. ase instance COFITZ_GDLF_EDC_1 is category_of_reuse = category_of_property = real_and_personal; type_of_property = olution is base= [ Medical Center, CO' ]; round= [ '1995' ]; source= [ 'BRAC95-COFITZ' ]; [ 'Golf course was operated by the City of' ]; contd_1 [ 'Aurora as a public course until later' ]; contd 2 [ as a bio-science research park.' ] contd 3 = [ 'Key issues reuolued around the sprinkler' ]; contd_4 [ installed below ground and who would' ]; contd 5 [ the grounds during winter ]; contd_6 [ 'after staff left and before LRA could' ]; contd 7 [ 'operate to Difficulties were' ]; contd 8 = [ 'encountered in whether the' ]; contd_9 [ 'sprinkler was HAF/AF and its ualue. LRA' ]; contd_10 = [ 'argued zero cost based on disposal cost' ]; contd_11 = [ 'to the if not giuen free to LRA.' ]; These two examples clearly demonstrate the benefits a corporate memory system can provide. Even though humans must formulate a solution, being able to address current problems with the knowledge of how attempts to solve similar problems in the past were successfully or unsuccessfully handled provides a great starting point for finding new solutions. Base Realignment and Closure is a domain that is not well defined, or where numerous exceptions exist. It is a domain where, versus computers, humans can still more efficiently adapt past experiences to solve problems. It is also a domain where there are not enough experts, but there is some experience. Therefore, the usefulness of a demonstration CBR corporate memory system in a domain such as BRAC shows that the above mentioned criteria for when to apply a CBR corporate memory system holds true. 79


APPENDIX C BASE REALIGNMENT AND CLOSURE PROCESS OVERVIEW Forts and ports have been created and closed throughout our military's history; however, Base Realignment and Closure, or BRAC, is a fairly new process. It is the Congressionally mandated closing or realigning of Department of Defense bases to reduce the military infrastructure due to federal budget reductions and the changing, global military mission. To date, BRAC Acts have been passed by Congress in 1988, 1990, 1993, and 1995. However, the rounds of BRAC are most likely not over as the Department of Defense has recently asked Congress for additional rounds in 2001 and 2007. The actual implementation of the BRAC Acts at a military installation is called the reuse implementation process. This process begins with the official announcement of a base for closure or realignment and ends with the base's conversion from military to civilian uses. The process can be viewed as three principal phases that are conducted concurrently: base-wide reuse planning, disposal decision making, and parcel-by-parcel 80


decision implementation. The following are summaries of the three phases {BRIM 1995} Phase One: Base-Wide Reuse Planning During this phase, the Local Redevelopment Authority {LRA} is established, typically by the community impacted by a base in their area realigning or closing. The LRA will coordinate and negotiate with the military on behalf of the community. During this phase, the LRA develops its redevelopment plan which identifies how the LRA plans to acquire parcels of land, buildings, and personal property. For example, a closing base may have a hospital that the LRA may plan to purchase and then lease to a commercial activity to generate revenue, or the LRA may plan to acquire the hospital as a public donation for the benefit of a state university's medical school. The LRA must complete its redevelopment plan and receive approval from the military before the next phase when the military makes decisions on how the base will be conveyed to the community. While the LRA is developing a redevelopment plan, the military conducts an environmental impact analysis, engages in environmental clean-up and compliance related activities, and identifies natural and cultural resources that should be protected. 81


Phase Two: Disposal Decision Making This phase includes activities associated with the military's decision process on how the base will be conveyed to the LRA based upon the LRA's redevelopment plan. The President has directed that the military cooperate with communities to close bases in a manner that meets closure timelines while also providing maximum support for redevelopment plans. The military will typically make one or more Disposal Record(s) of Decision (RODs) or similar decision documents to announce to the general public how it plans to convey the base to the LRA or others. Phase Three: Parcel-by-Parcel Decision Implementation After final disposal decisions have been issued by the military, this last phase begins and does not end until all available real and personal property is conveyed. The actual conveyance of real property cannot occur until the property has been cleared by the Environmental Protection Agency. For property with ongoing environmental clean-up efforts, leases may be used to achieve prompt reuse. 82


APPENDIX D OTHER IMPORTANT BRAC PROBLEM AREAS The BRAC process is tremendously large, complex, and as a whole, is an enormous planning and scheduling problem. The military needs to eliminate infrastructure that is no longer needed to support a shrinking military. This reduction needs to be done as rapidly as possible to achieve cost savings. On the other hand, the community seeks rapid reuse and redevelopment with quick economic recovery as the goal. These two objectives often conflict, but the military does recognize the need to close bases in manner that achieves rapid closure while still meeting community needs whenever possible. In addition to the overall problem of time constraint and priority conflicts between the military and community, the following are more specific problems that could benefit from the application of computer science. 83


Disseminating/Retaining BRAC Community Redevelopment Knowledge Much about the way the military goes about executing the BRAC process at a base depends on the community's desire or plan for redevelopment of the base. The DoD guideline is for the creation of a redevelopment plan to take the community approximately one year and a half. Speeding up this planning process would significantly contribute to both the military and community benefiting from an accelerated closure. A "corporate knowledge" system similar to the one identified above for the military, but geared toward BRAC communities, would certainly help in this area. Such a system would likely have commercial value if future rounds of BRAC do occur. Environmental Clean-up Advising Environmental issues are one of the most critical elements of the BRAC process. Numerous environmental actions such as studies, actual clean-up or remediation, and approvals must occur before property may be conveyed (or even leased) to the community for redevelopment. Determining the appropriate base line environmental 84


tests, and the need for additional testing and any cleanup requirements and courses of action could be assisted through consideration of similar cases from other BRAC sites. Although not under the realm of Knowledge Based Systems, other artificial intelligence methods could certainly aid in planning and scheduling the numerous assessments that occur simultaneously or sequentially predicated on earlier environmental test results. Testing and clean-up actions could be weighted based on the level of threat, need for site conveyance/reuse, and other factors. Real Property Disposal Planning Real property includes land, buildings, and other permanent structures. At a BRAC site, the military seeks to convey real property to the community for cost savings and perhaps will prioritize which real property to dispose of based on factors such as: current occupancy or use, operational costs, maintenance requirements, amount of associated personal property, environmental and safety factors, and so on. On the other hand, the community seeks to acquire real property to maximize their redevelopment opportunity. The community may request the 85


military prioritize the turnover of real property based on suitability for commercial reuse, condition, location, and so forth. The process to prepare parcels of real property for conveyance are extensive and very time consuming, involving: environmental studies and clean-up; inventories of related personal property; real and personal property valuation; license, lease, purchase, or other conveyance negotiations; document creation; and much more. A planning and scheduling system could be created, allowing various weights to be placed on military and community decision/priority factors like those listed above. Such a system would aid the military in developing a real property processing schedule that meets military requirements and provides the best possible support for community reuse needs. Mission Closure Planning Similar to the above problem is planning and scheduling the closure of operational areas, especially at BRAC sites that will fully close. Activities on a base support each other, but must ultimately close. Careful consideration and planning is needed to schedule activities for closure to ensure activities have the 86


necessary closure support. For example, the Directorate of Information Management (DOIM} needs Directorate of Logistics (DOL} support to account for personal property, provide equipment maintenance, and physically move equipment for turn-in or transfer. Likewise, DOL needs DOIM support to maintain automated property accountability systems, provide a local network, phones, and so on. There are literally hundreds of such activities typically on a base, and the relationships between activities occur at multiple levels, have complex dependencies, and varying degrees of importance. A system that can help sort out and plan in this area would be of tremendous benefit. 87


ANNOTATED BIBLIOGRAPHY [Aamodt 1994] Aamodt, Agnar, and Plaza, Enric, Based Reasoning: Foundational Issues, Methodological Variations, and System Approaches". Artificial Intelligence Communications, Vol. 7, No. 1, 1994, pp39-59. This article provides an overview of the foundational issues of CBR and discusses some of the leading methodologies and provides examples through some CBR systems. The article also provides a summary of methods for case retrieval, reuse, solution testing, and learning. Case-based reasoning methods are discussed as one type of reasoning and learning method within an integrated system architecture. [Adelman 1992] Adelman, Leonard, Evaluating Decision Support and Expert Systems. NY: John Wiley & Sons, Inc., 1992. The author of this book admits that decision support and expert systems hold great promise, but claims progress has been slowed in these areas due to a technology driven development of systems, rather than a requirements driven approach. The author focuses on ways to incorporate evaluation into development and then discusses three development methods: subjective, technical, and empirical. The author concludes by addressing how to manage the evaluation process. [Barrett 1988] Barrett, Mike, Expert Systems in Business: A Practical Approach. Chichester, England: Ellis Horwood Limited, 1988. This book is geared toward corporate managers, MIS specialists, and experts system developers. The book is a practical guide on how to use expert systems to leverage a business advantage. The author clearly outlines when to apply expert system 88


technology and what steps managers should take to integrate such systems into their business operations. [Beerel 1986] Beerel, Annabel C., The Design and Use of Expert Systems. NY: John Wiley & Sons Limited, 1986. The primary focus of this work is to serve as a guide to exploiting expert systems in business. It identifies what business managers can and should do with expert systems in their own company. Special attention is given to business opportunities and software challenges. [Beerel 1987] Beerel, Annabel C., Expert Systems: Strategic Implications and Applications. NY: Ellis Horwood Ltd, 1987 .. This is a book about the strategic implications and applications of expert systems. It's focus addresses the current movement toward a knowledgebased society, the new knowledge worker, and the fundamental role that expert systems, and information technology in general, will play in "harnessing inherent knowledge to the strategic advantage of all organizations." [Belgum 1990] Belgum, Erik, Artificial Intelligence: Great Mysteries & Opposing Viewpoints. San Diego: Greenhaven Press, 1990. This book is about AI, but is part of a book series that covers many other topics. As a result, this book provides a_ very low level introduction to AI. The book gives a very brief history of AI, discusses intelligence, natural language programming, and the future of AI. [BRIM 1998] Assistant Secretary of Defense for Economic Security, Base Reuse Implementation Manual (BRIM). Alexandria, VA: Office of the Assistant Secretary of Defense for Economic Security, 1998. This manual was prepared by the Office of the Assistant Secretary of Defense for Economic Security [OASD(ES)], in cooperation with the Military Departments, and the Office of the Secretary of Housing and Urban Development. Base Realignment and closure is governed by 32Code of Federal 89


Regulations (CFR), parts 90 and 91. The BRIM provides supplemental guidance for those carrying out BRAC. In addition, the BRIM provides some common-sense approaches and general practices for the Military Departments to follow during the BRAC process. Users of the manual are encouraged to adapt the guidance in the BRIM to their own installation-specific circumstances. [Bundy 1990] Bundy, ALan, Ed., Catalogue of Artificial Intelligence Techniques. Berlin: Springer-'Verlag, 1990. This work catalogues artificial intelligence techniques (from 21/z-D Sketch to Vowel Quadrilateral) to "promote interaction between members of the AI community" by providing access to a common, extensional definition of the field to "promote a common terminology, discourage the reinvention of wheels, and act as a clearing house for ideas and algorithms." [CGPPD 1995] Assistant Chief of Staff for Installation Management, A Commander's Guide to Personal Property Disposal (CGPPD) Alexandria, VA: Office of the Assistant Chief of Staff for Installation Management, 1995. This guide provides guidance to commanders at Base Realignment and Closure sites on the daunting task of inventorying and disposing of all the personal property on the BRAC installation. Personal property is all property other than buildings, land, and other fixed assets and realestate. So, personal property includes fighter aircraft to a thumb tack. The guide addresses all issues concerned with the inventorying, conveyance, turn-in, and transfer of such property. [Cleal 1988] Cleal, D.M. and Heaton, N.O., KnowledgeBased Systems: Implications for Human-Computer Interfaces. NY: John Wiley & Sons, 1988. The authors present the case that those in AI need the knowledge of those working in human factors (and vice versa) The authors describe KBS and AI, and KBS and human factors. The authors then describe how the fields overlap and 90

PAGE 100

investigate the potential for better expert systems which such an overlap can create. [Diaper 1989] Diaper, Dan, Knowledge Elicitation: Principles. Techniques and Applications. Chichester, England: Ellis Horwood Limited, 1989. The author addresses the knowledge elicitation bottleneck encountered with knowledge based systems in this book. The author focuses on Task Analysis as the major methodological approach available to aid in overcoming the knowledge elicitation bottleneck, but the book contains many other task analysis methods that are useful in system design and development. [Dieng 1996] Dieng, Rose and Giboin, Alain, "Building of a Corporate Memory for Traffic Accident Analysis". Paris, France: Universite' Rene' Descartes, 1996. The article presents the authors' study and comparison of the integration of models of expertise obtained from seven experts in accidentology. The article presents the elicitation protocol used to gather the experts knowledge, and the generic models and tools used for knowledge modeling. The authors conclude by discussing the results of the experiment from a knowledge capitalization viewpoint. [Fischler 1987] Fischler, Martin A., Intelligence: The Eye. the Brain. and the Computer. Menlo Park, CA: Addison-Wesley Publishing, 1987. The book discusses the operation of the human brain, the concept of intelligence, the nature of cognitive and perceptual capabilities of people and machines, algorithms used to model intelligent behavior, and the present-day and ultimate limits of machine performance. [Freedman 1994] Freedman, David, Brainmakers: How Scientists are Moving Beyond Computers to Create a Rival to the Human Brain. NY: Simon.& Schuster, 1994. This book is a documentary of sorts that provides an introduction to the work of various scientists around the world whose 91

PAGE 101

work is aimed at creating devices that are more like living brains than computers that includes creations such as: a robot farm where robots are breed for intelligence, a collection of chemicals that act like a primitive life form and can recognize patterns, a machine that experiences human-like brain waves and mental disorders, and a new form of computer chip out of living brain cells. [Gaines 1988] Gaines, B.R. and Boose, J.H., Knowledge Acquisition for Knowledge-Based Systems. NY: Academic Press, 1988. This book is a collection of papers from the AAAI workshop in November, 1986. The collection contains six overview/summary papers and contains other papers that address cognition and expertise, interactive interviewing tools, learning, reasoning with uncertainty, analysis of knowledge structures, and knowledge representation. [Garnham 1987] Garnham, Alan, Artificial Intelligence: An Introduction. NY: Routledge & Kegan Paul, 1987. The title gives this one away. In addition to the basics, the book does specifically address the application of artificial intelligence in the fields of vision, language, and learning. [Gelernter 1994] Gelernter, David, The Muse in the Machine: Computerizing the Poetry of Human Thought. NY: The Free Press, 1994. In this work, Gelernter emphasizes the role of logic in thinking while also stressing the importance of emotions in our daily thought processes. The book analyses both childlike thought and ancient thought, and provides a model for introducing emotion into the computer. [Gill 1986] Gill, Karamjit S., Artificial Intelligence for Society. NY: John Wiley & Sons, 1986. This is a collection of essays on the social aspects of artificial intelligence. The book is organized around the following AI issues: problems and perspectives, philosophical 92

PAGE 102

issues, culture and the arts, social issues, technology and education, and technology and applications. [Groendijk 1993] Groenjijk, Cees and Oskamp, Anja, Recognition and Strategy Classification". Association for Computing Machinery, Vol.93, No.6, 1993, pp125-132. The article discusses the benefits of obtaining automated reasoning control knowledge (search and strategy) from precedents which resemble the current fact situation. The model combines rule-based strategy with a case-based reasoner. The case-based reasoner is used when the system comes across concepts or terms that cannot be resolved given the facts of the current case and the rule-sets provided. Neural networks to find similar cases. The model allows the suggestion of a similar case with applicable control strategy in a very early stage of problem solving. [Hage 1993] Hage, Jaap, Reason-Based Logic: A Low Level Integration of Rule-Based Reasoning and CaseBased Reasoning". Association for Computing Machinery, Vol.93, No.6, 1993, pp30-39. This article provides an introduction to reason-based logic (legal reasoning) The authors argue that a conclusion in legal reasoning comes down to collecting the reasons that support and oppose a conclusion, and then weighting them. The article then reasons how projections can then be made as to whether the reasons for or against a conclusion will prevail. The article also addresses meta-level reasoning about the use of rules in concrete legal cases. [Hand 1993] Hand, D.J., Artificial Intelligence Frontiers in Statistics. NY: Chapman & Hall, 1993. This work is a collection of papers presented at the Third International Workshop on Artificial Intelligence and Statistics, Fort Lauderdale, Florida, in January, 1991. The collection of papers is divided among the general headings of: statistical expert systems (including much attention to commercial applications), belief networks, learning, neural networks, text manipulation, and other 93

PAGE 103

areas such as a paper on statistical and artificial intelligence approaches to combining the probability judgments of experts. [Hennessy 1992] Hennessy, D.H., and Hinkle, D., "Applying Case-Based Reasoning to Autoclave Loading". IEEE Expert, Vol. 7, No. 5, 1992, pp21-26. The paper provides the authors' findings associated with the casebased reasoning system, CLAVIER, developed and implemented as part of Lockheed's aircraft manufacturing operations. CLAVIER takes information on composite aircraft parts that need to be created and advises users on what past autoclave loads worked that were most similar to the current batch of parts. [Hickman 1989] Hickman, Frank R., Analysis for Knowledge-Based Systems: A Practical Guide to the KADS Methodology. NY: John Wiley & Sons, Inc., 1989. This book is a guide to the KADS methodology as the title clearly states. In addition to an introduction, overview, and history of KADS, the book provides detail on the KADS framework for analysis, provides a model for knowledge analysis, and addresses knowledge capture and model building. The bo9k concludes with a case study on building an interpretation model using generic models and a case study on constructing an interpretation model from primitives. [Hogan 1997] Hogan, James P., Mind Matters: Exploring the World of Artificial Intelligence. NY: The Ballantine Publishing Group, 1997. The author chronicles the efforts of the artificial intelligence community beginning with Aristotle and moving on to clocks and calculators, and still on to computers, neural nets and perceptrons, and robot vehicles probing Mars. The book is loaded with references to theories, methodologies, and projects. [Hunt 1986] Hunt, V. Daniel, Artificial Intelligence & ExPert Systems Sourcebook. NY: Chapman & Hall, 1986. 94

PAGE 104

This work contains a brief introduction to artificial intelligence and expert systems. The remainder of the book contains definitions related to artificial intelligence and expert systems. Finally, the book provides points of contact and an explanation of acronyms. [Joshi 1996] Joshi, Suneela R. and McMillan, William w, "Case Based Reasoning Approach to Creating User Interface Components". Association for Computing Machinery, Vol.96, No.4, 1996, pp81-82. This article proposes using CBR as an interactive software reuse tool. Software developers can save time and money by reusing code for user interface tools such as windows, menus, icons, dialogues, and such. A CBR based interactive software reuse tool can help a developer select interface components for reuse that are appropriate for the developer's project. The authors propose that such a CBR tool will eliminate coding and researching for the best suitable interface components for a software application. [Keller 1987] Keller, Robert, Expert System Technology: Development & Application. Englewood Cliffs, NJ: Yourdon Press, 1987. Expert System development methodology, selecting expert system applications, a study in application selection, an example ES, a case study in structured analysis, and other general AI topics such as frames, inference, and knowledge. [Kemmerer 1989] Kemmerer, Richard A., Proceedings of the ACM SIGSOFT '89/Third Symposium on Software Testing, Analysis. and Verification CTAV3), Key West, Florida. December 13-15. 1989. NY: ACM Press, 1989. These proceedings cover theoretical models, tools, specification-based approaches, empirical studies, data flow testing and tasking supervisor testing, and integrating techniques among other topics related to software testing, analysis, and verification. 95

PAGE 105

[Kitano 1992] Kitano, H; Shibata, A; Kajihara, J; and Sate, A, "Building Large-Scale and Corporate-Wide CaseBased Systems". In Proceedings of AAAI-92. Cambridge, MA: MIT Press, 1992. This article discusses the authors' work at NEC regarding the creation of SQUAD, a case-based reasoning system that collects, organizes, and distributes information about corporate software quality control/trouble shooting. [Kitano 1996] Kitano, Hiroaki, and Shimazu, Hideo, "The Experience-Sharing Architecture: A Case Study in Corporate-Wide Case-Based Software Quality Control". In Case-Based Reasoning: Experiences, Lessons, & Future Directions. Cambridge, MA: MIT Press, 1996. This article describes CASEMETHOD, the authors' methodology for creating large-scale case-based reasoning systems. The authors provide a case study involving SQUAD, a casebased reasoning system that collects, organizes, and distributes information about corporate software quality control/trouble shooting for NEC. [Kolodner 1984] Kolodner, Janet, Retrieval and Organization Strategies in Conceptual Memory: A Computer Model. Northvale, NJ: Erlbaum, 1984. This work contains Janet's work on developing an automated representation of the model for human memory retrieval and organizational. It includes information on CYRUS, a case-based reasoning system that contains information on the travels and meetings of Cyrus Vance, a former US Secretary of State. [Kolodner 1988] Kolodner, Janet, ed., Case-Based Reasoning: Proceedings of a Workshop on Case-Based Reasoning: Holiday Inn, Clearwater Beach, Florida, May 10-13, 1988. San Mateo, CA: Morgan Kaufmann Publishers, 1988. This work is a collection of papers from proceedings of a workshop on case-based reasoning that occurred in Clearwater Beach, FL, in May 1988. Papers included in the work are those on: adaptation and repair, application domains, application support, CBR background, design problem solving, diagnosis, domain-independent 96

PAGE 106

inferences, explanation-based CBR, exploratory programs, feasibility and uses of CBR, indexing and retrieval, integrating CBR and other problem solving methods, integrating execution and planning, learning from cases, mapping, open worlds, opportunistic problem solving, planning, psychological investigations, case representation, and tasks for a CBR reasoner. [Kolodner 1993] Kolodner, Janet L., Case-Based Reasoning. San Mateo, CA: Morgan Kaufmann Publishers, 1993. This book serves as an excellent introductory text to CBR. A thorough background on CBR is provided that includes reasoning using cases and the cognitive model. The text continues with discussions on the case library (representing and indexing cases), case retrieval, and case adaptation, interpretation, and evaluation. The work ends with an example of building a case-based reasoner and includes a case library of CBR systems. [Koulopoulos 1997] Koulopoulos, Thomas, "Knowledge Management: Toward Creating the 'Knowing Enterprise'". Boston, MA: The Delphi Consulting Group, 1997. This white paper addresses the evolution of knowledge management from simple corporate memories to a true knowledge management system that can help achieve corporate instinct. The paper also discusses how to build a seamless knowledge management solution and explains information life cycle management. [Kowalski 1991] Kowalski, Andrzej, "Case Based Reasoning and the Deep Structure Approach to Knowledge Representation". Association for Computing Machinery, Vol.91, No.6, 1991, pp21-30. This article discusses how CBR legal expert systems are superior to similar, purely rule-based systems. The authors claim one of the most important advantages CBR has over purely rule-based systems is that CBR systems more realistically portray the reasoning process of a lawyer. The article demonstrates that the deep structure approach to knowledge representation and a relatively inexpensive and 97

PAGE 107

commercially available CBR shell can be used to build a CBR legal expert system in case law which: dynamically draws its conclusions without the benefit of specific legal rules; adapts to most changes in law simply through the entering of new cases; adjusts its outcomes depending on the "home" jurisdiction/locality; overcomes the difficulties of representing legal knowledge; and operates at a high level of expertise. [Kriz 1986] Kriz, J., Knowledge-Based Expert Systems in Industry. NY: John Wiley & Sons Limited, 1986. The book addresses the applications of artificial intelligence in knowledge-based systems. The book contains reports from various academia and industry experts with the following being addressed: current research, future trends, and knowledge-based systems for configuration, planning, and diagnosis. In addition, expert system shells and high level programming languages are covered as well. [Kuhn 1998] Kuhn, Otto and Abecker, Andreas, "Corporate Memories for Knowledge Management in Industrial Practice: Prospects and Challenges". Graz, Austria: Institute for Information Processing & Computer Supported New Media, 1998. The paper compares and summarizes the authors' experiences with three case studies on corporate memory systems for supporting various aspects in the product life-cycles of three European companies. The authors sketch a general framework for a development methodology and architecture of a corporate memory system. [Lassez 1991] Lassez, Jean-Louis, ed., Computational Logic: Essays in Honor of Alan Robinson. Cambridge, MA: MIT Press, 1991. This book is a collection of essays that address the main aspects of computational logic, a field for which Alan Robinson is largely credited with the conception. Major topics included in the work are: inference, equality, and logic programming. 98

PAGE 108

[Leake 1996] Leake, David B., ed., Case-Based Reasoning: Experiences. Lessons. & Future Directions. Cambridge MA: MIT Press, 1996. The book contains a collection of writings from various experts from the CBR field. The contributors provide issue-oriented case studies of experiences with particular projects to provide a view of the principles of CBR. Key issues covered are: indexing, retrieval, case adaptation, evaluation, and application of CBR methods. [Marik, 1993] Marik, Katharina and Wrobel, Stefan, Knowledge Acquisition and Machine Learning: Theory. Methods. and Applications. NY: Academic Press, 1993. This book reports work on a series of systems oyer ten years. This book focuses on knowledge-intensive, logic-based learning and the current debate on inductive logic programming. In addition, the book gives detailed attention to the tools that assist with modeling an application domain -tools that: structure predicates, discover rules, and help with knowledge revision and concept learning). Finally, experience with running the various systems on applications is presented. [Murbach 1993] Murbach, Ruth and Nann, Eva, "Similarity in Harder Cases: Sentencing for Fraud". Association for Computing Machinery, Vol.93, No.6, 1993, pp236-244. This article describes experiments in using FXS, a program designed to retrieve similar cases based on their degree of salience, in the legal process of sentencing where cases are dependent on various legitimate principles, objectives, and factors which relate to the offender, the victim, and the act within its social context. The authors model uses a salience coefficient to measure similarity based on relative high or low frequency of factors in their local context. [Nikolopoulos 1997] Nikolopoulos, Chris, Expert Systems: Introduction to First and Second Generation and Hybrid Knowledge Based Systems. NY: Marcel Dekker, Inc., 1997. This book was written as an undergraduate text to provide 99

PAGE 109

an introduction to expert systems. The work addresses: general, underlying expert system concepts and theories; knowledge representation schemes; knowledge acquisition, verification, and validation; handling uncertainty; machine learning and the knowledge elicitation bottleneck; the connectionist approach and neural networks; and hybrid systems. [Penrose 1989] Penrose, Roger, The Emporer's New Mind: Concerning Computers, Minds, and the Laws of Physics. NY: Oxford University Press, 1989. This book covers many, many topics. Written with humor and for "informed laymen", the book hits on relativity theory, quantum mechanics, cosmology, a fractal-like structure called the Mandelbrot, complex numbers, Turing machines, formal systems, phase spaces, Hilbert spaces, black holes, white holes, and the structure of the brain. Especially related to AI is Penrose's strong argument against "strong AI", which according to Penrose is belief that the human mind is equivalent to nothing more than just a collection of tiny wires and switches. [Pinkerton 1987] Pinkerton, J. M. M., Information Technology: An Overview. NY: John Wiley & Sons Limited, 1986. This book describes the fundamentals associated with computers and our need to communicate. The book is written as an introductory work and covers current information technology methods and addresses the future directions of information technology as well. [Rich 1991] Rich, Elaine and Knight, Kevin, Artificial Intelligence. NY: McGraw-Hill, 1991. This book is designed as a college text to provide an introduction to the broad field of artificial intelligence. The book starts with basics such as problems, problem spaces and search, knowledge representation, predicate logic, and rules. In addition, the book covers symbolic reasoning under uncertainty, statistical reasoning, and weak and strong slot-and-filler structures. The book finishes up with advanced topics on game playing, planning, 100

PAGE 110

understanding, natural language processing, parallel and distributed AI, learning, common sense, and expert systems. [Ringland 1988] Ringland, G. A. and Duce, D. A., Approaches to Knowledge Representation: An Introduction. NY: John Wiley & Sons, Inc., 1988. This book originated from a series of lectures on knowledge representation given by the authors at Rutherford Appleton Laboratory. The book explains and analyzes a wide range of knowledge representation approaches, including logic, semantic networks, frames, and rule-based systems. The book also covers how humans appear to represent knowledge and addresses some advanced topics such as: representing time, meta-knowledge, conceptual graphs, computational tractability, and functional approaches. [Rolston 1988] Rolston, David W., Principles of Artificial Intelligence and Expert Systems Development. NY: McGraw-Hill Book Company, 1988. This detailed, yet very "readable" book is geared to developers, more than researchers. The book covers is good detail many topics such as: AI problem solving concepts; formal and informal knowledge representation; formal logic; uncertainty; explanation facilities; the expert system development process; knowledge acquisition; expert system tools; inference based on formal logic; and hybrid systems. The book includes appendixes on Lisp and Prolog programming. [Sage 1991] Sage, Andrew P., Decision Support Systems Engineering. NY: John Wiley & Sons, Inc., 1991. The detailed assessment of decision support systems presented in this book is directed at decision support system designers as well as system users. The book begins with an overview of system components and basics. The author covers database management systems, model-based management systems, and dialog generations systems. However, the primary focus of the book is teaching when a decision support system is feasible and practical. 101

PAGE 111

[Savory 1987] Savory, S.E., Expert Systems in the Organisation. Chichester, England: Ellis Horwood Limited, 1987. This book is written more for business managers than for more technical readers. The book provides a high level introduction to expert systems while focusing on the benefits of such systems and when such systems are applicable and economical. [Schank 1994] Schank, Roger C., Kass, Alex, and Riesbeck, Christopher K., Inside Case-Based Explanation. Hillsdale, NJ: Lawrence Erlbaum Assoc., 1994. In this work, Schank and his partners take on the explanationconstruction problem. The author points out that this is a computationally intractable problem, admitting that building a completely new explanation from scratch every time would leave us with nothing else to do. By using the case-based paradigm, the authors suggest the main steps to the process of explaining an anomaly becomes: retrieve an explanation that might be relevant to the anomaly, evaluate whether the retrieved explanation makes sense when applied to the current anomaly, and if the explanation doesn't fit the current anomaly perfectly, then adapt the explanation to make it fit better. The book describes some detailed theoretical work and ends with discussions of implementation. [Scott 1992] Scott, Carlisle A. and Klahr, Philip, Innovative Applications of Artificial Intelligence 4. Menlo Park, CA: AAAI Press, 1992. This book contains proceedings of the IAAI-92 conference that includes papers written by various authors on the application of artificial intelligence in the following domains: customer service, data analysis, finance, industrial, regulatory, routing, and software development. [Silverman 1995] Silverman, Barry G., "Intelligent Multimedia Repositories (IMRs) for Reuse-Supported Work: An Empirical Study'. Washington, D.C.: George Washington University Institute for Artificial Intelligence. This research explores whether the use of multimedia and 102

PAGE 112

intelligent agents foster the reuse of artifacts from a repository. A study was conducted with 33 professional subjects in a software project estimation repository. Performance and reaction data were collected on the purpose, role, usefulness, impact, and importance of 14 media/artifact categories. Specific lessons learned are offered for the design of reuse repositories, the use of multimedia, and the role for intelligent agents. [Slater 1986] Slater, Philip E., Building Expert Systems: Cognitive Emulation. NY: John Wiley & Sons Limited, 1986; This book addresses the question, "To what extent can an expert system emulate human thinking?" To achieve his goal, the author addresses the issue of the assessment of the viability and implications of a cognitive approach to knowledge engineering. [Smith 1993] Smith, Suzanne and Kandel, Abraham, Verification and Validation of Rule-Based ExPert Systems. London: CRC Press, 1993. The authors of this book adapt the concepts of verification and validation for traditional software and apply it to the domain of expert systems. Where traditional methods do not apply to expert systems, the authors identify new methods. Furthermore, the authors provide a complete set of techniques and tools to provide a formal, automated means of verifying rule-based expert systems. [Tansley 1993] Tansley, D. S. and C. C. Hayball, Knowledge-Based Systems Anaylsis and Design: A KADS Developer's Handbook. NY: Prentice Hall, 1993. As the title suggests, this truly is a developer's,handbook. The author enlightens the reader on the hazards of applying ill-structured techniques {usually based on evolutionary prototyping) and then dives into the KADS methodology. The book serves as an introduction and full explanation of KADS, but is most of all a practical guide to the use of KADS methods. 103

PAGE 113

[Thompson 1997] Thompson, Valerie, "Corporate Memories", NY: BYTE Magazine, Sep 1997. The author of this article provides a brief review of the history of corporate memory, touches on the basic concepts of corporate memory, and addresses some of the commercially successful applications of corporate memories. The article ends with a discussion of the future of corporate memory and its overall role in corporate efforts to manage their collective knowledge. [Vossos 1991] Vossos, George, "An Example of Integrating Legal Case Based Reasoning with Object-Oriented RuleBased Systems: IKBALS II". Association for Computing Machinery, Vol.91, No.6, 1991, pp31-41. This paper discusses extending an existing legal objectoriented/rule-based system to incorporate CBR. The authors use IKBALS II as an example. IKBALS is a legal knowledge-based system that performs statutory interpretation in the area of accident compensation under the Accident Compensation Act, 1985, Victoria. [Walters 1988] Walters, John and Nielson, Norman R., Crafting Knowledge-Based Systems: Expert Systems Made Realistic. NY: John Wiley & Sons, 1988. This book is a detailed guide for developers of knowledge-based systems. The book begins by discussing planning and designing a knowledge-based application, including information of forming the development team, gathering requirements, conducting a feasibility study, and acquiring knowledge. The book continues by discussing how to craft a knowledge-based application (prototyping, evaluation, pilot applications, and project scheduling) and ends with a discussion of knowledge representation. [Warnier 1986] Warnier, Jean-Dominique, Computers and Human Intelligence. Englewood Cliffs, NJ: Prentice-Hall, 1986. This book begins with a history and current assessment of computers and computing science. The book goes on to discuss common sense and the logic of computing, computers as a substitute for humans, and 104

PAGE 114

computers as a human tool. The book concludes with the author's views on the teaching of computing. [Waterworth 1986] Waterworth, J. A., Speech and Language-Based Communication with Machines. NY: Ellis Horwood Ltd, 1987. The author focuses on research efforts on the psychology of speech perception, and production and developments in machine speech synthesis and recognition. However, the author also addresses the associated artificial intelligence work on dialogue design and knowledge representation. [Watson 1997] Watson, Ian, Applying Case-Based Reasoning: Techniques for Systems. San Francisco: Morgan Kaufmann Publishers, 1997. This book is geared toward a broad audience that includes most all academic and industry fields which CBR may benefit. The book contains very little theory, primary focusing on case studies involving CBR systems in the commercial sector. The author provides a brief history of CBR, compares CBR to other techniques, and then jumps right in to industrial applications of CBR. A detailed list and comparison of CBR software tools is provided and then an example is provided for building a diagnostic case-base. [Yamaguti 1993] Yamaguti, Takahira and Kurematsu, Masaki, "Legal Knowledge Acquisition Using Case-Based Reasoning and Model Inference". Association for Computing Machinery, Vol.93, No.6, 1993, pp212-217. The authors state that, while CBR is good at avoiding the knowledge elicitation bottleneck, a case structure acquisition bottleneck has emerged for CBR. The authors propose a reduction in the case structure acquisition bottleneck through a framework for CBR where case structures are improved dynamically. This is made possible through the use of theoretical term generation (the operation to induce a useful concept and give it a name, or obtain new descriptors to define a problem) 105

PAGE 115

[Yang 1993] Yang, Soon-Ae; Robertson, Dave; and Lee, John, "KICS: A Knowledge-Intensive Case-Based Reasoning System for Statutory Building Regulations and Case Histories". Association for Computing Machinery, Vol.93, No.6, 1993, pp254-263. The authors of this article claim that knowledge based systems related to statutory building regulations during the past decade have not taken advantage of existing case histories. This article proposes a CBR system which can not only be used for retrieval and maintenance of building regulations, but also for case histories. The authors argue that including case histories is important because deviation from statutory requirements allow deviation for exceptional or unexpected circumstances. [Ziman 1989] Ziman, John, Benefits and Risks of Knowledge-Based Systems. NY: Oxford University Press, 1989. This work is produced by The Council for Science and Society, a registered charity with the object of "promoting the study of, and research into, the social effects of science and technology, and of disseminating the results thereof to the public." In particular, the book "aims to alert the public to the potential benefits, and the possible dangers, of a new form of advanced information technology: knowledge-based systems." The book provides an explanation of how knowledge-based systems work and identifies some applications of KBS. The book describes the benefits of KBS, and also sheds light on what the social critics of KBS have to say. The book ends with conclusions and recommendations. 106