1 Methods and Tools for Corporate Knowledge Management

Rose Dieng, Olivier Corby, Alain Giboin, Myriam Ribière

*INRIA Sophia-Antipolis, Projet ACACIA,

2004 route de Lucioles,BP 93,

06902 SOPHIA-ANTIPOLIS CEDEX, FRANCE,E-mail: {dieng,corby,giboin,ribiere}@sophia.inria.fr

Abstract.This article is a preliminary survey of some methods, techniques and tools aimed at managingcorporate knowledge from a corporate memory (CM) designer's perspective. In particular, it analyzes pro-blems and solutions related to the following steps: detection of needs of CM, construction of the CM, itsdiffusion (specially using the Internet technologies), its use, its evaluation and its evolution.

1

1.1

INTRODUCTION

Corporate Memory: Definitions

The objectives of knowledge management (KM) in an organization are to promote knowledgegrowth, knowledge communication and knowledge preservation in the organization (Steels, 93).Knowledge management is a very complex problem and can be tackled from several viewpoints:socio-organizational, financial and economical, technical, human, and legal (Barthès, 1996).

There is an increasing industrial interest in the capitalization of knowledge (i.e. both theoreticalknowledge and practical know-how) of groups of people in an organization, such groups beingpossibly dispersed geographically. In (Van Heijst, Van der Spek, and Kruizinga, 1996)«corporatememory» is defined as an«explicit, disembodied, persistent representation of knowledge andinformation in an organization». For example, it may include knowledge on products, productionprocesses, clients, marketing strategies, financial results, plans and strategical goals, etc. (Nagen-dra Prasad and Plaza, 1996) define corporate memory as «the collective data and knowledgeresources of a company including project experiences, problem solving expertise, design ratio-nale, etc»: it may include databases, electronic documents, reports, product requirements, designrationale, etc. It is a «repository of knowledge and know-how of a set of individuals working in aparticular firm» (Euzenat, 1996) and its building relies on the«will to preserve, in order to reusethem later or the most rapidly, reasonings, behaviours, knowledge even in their contradictionsand with all their variety»(Pomian, 1996). Knowledge capitalization is the process which allowsto reuse, in a relevant way, the knowledge of a given domain, previously stored and modelled, inorder to perform new tasks (Simon, 1996). The purpose is to«locate and make visible the enter-prise knowledge, be able to keep it, access it and actualize it, know how to diffuse it and better useit, put it in synergy and valorize it»(Grundstein, 1995).

Several kinds of knowledge can be found in a company:explicit ortacit knowledge (Nonaka,1991). Therefore, in any operation of knowledge capitalization, it is important to identify crucialknowledge to be capitalized (Grundstein and Barthès, 1996). It has an influence on the kind ofCM needed by the enterprise. This CM should help to support the integration of resources andknow-how in the enterprise and the cooperation by effective communication and active documen-tation (Durstewitz, 1994).

1

As noticed in (Nonaka, 1991; Van Engers, Mathies, Leget and Dekker, 1995), the knowledgechain consists of seven links: listing the existing knowledge, determining the required knowledge,developing new knowledge, allocating new and existing knowledge, applying knowledge, main-taining knowledge, disposing of knowledge. So, we can consider the building of the CM asrelying on the following steps (summed up in Figure 1): (1)Detection of needs in corporatememory, (2)Construction of the corporate memory, (3)Diffusion of the corporate memory, (4)Use of the corporate memory, (5)Evaluation of the corporate memory, (6)Maintenance and evo-lution of the corporate memory.

Enterprise ModelsBusiness process reengineeringKnowledge ServersInformation retrievalGroupwareDetectneedsBuildDistributeUseEvaluateMake evolveSourcesPersonsNotes, reports,manuals, guides,drawingsDocumentsDatabasesNaturePaper-based documentsvs electronic documentsFormal knowledgevs informal knowledgeKnowledge baseCase baseMethodsTechniquesKnowledge engineering techniquesCase-based reasoningAgentsCooperative creationDistributed memoryLinguistic analysisHypertextsFig.1: Corporate Memory Management

For each step, we will analyse some methodological or technical proposals offered by researchers.Let us notice that several kinds of publications can be found: survey on KM, analysis of types ofknowledge available in a company, reports of industrial experiments, proposal of a general archi-tecture for CM, thorough study of a particular technique such as some knowledge-processingtechniques stemming from artificial intelligence (AI) and used here for solving a peculiar problemunderlying computational CM building. The variety of research topics possibly involved in CMmanagement is illustrated by Figure 1. Clearly, this complex problem has at least organizationalaspects to be tackled, and technical aspects to be solved. According to (Kühn and Abecker, 1997),computer scientists concerned by the use of Information and Communication Technology forKM support tend to ignore the specific requirements and constraints for successful knowledgemanagement in industrial practice while specialists in KM often treat only roughly the aspects ofcomputer support. Therefore, building a CM requires a multidisciplinary approach.

2

1.2Corporate Memory Industrial Needs

An enterprise is not only a unit of production of goods or services conform to the expectations ofclients, in the best conditions of cost, deadline and quality, but it is also a knowledge productionunit (Grundstein, 1995). The nature of the needed CM and the efforts needed for building it maydepend on the size of the company (cf. wide-sized groups vs small-sized and medium-sizedfirms). The motivations can be various: (a) to avoid loss of know-how of a specialist after hisretirementor mutation, (b) to exploit the experience acquired from past projects, and to keepsome lessons from past, in order to avoid to reproduce some mistakes, (c) to exploit theknowledge map of the company for the corporate strategy: a regular inventory of the firm know-how should improve the enterprise ability to react and adapt to change, (d) to improve informa-tion circulation and communication in the enterprise, (e) to improve learning of employees in theenterprise (new as old employees), (f) to integrate the different know-how of an organization.1.3

Knowledge in the Enterprise

Several typologies of knowledge in the enterprise were proposed in literature. They can be usefulto determine the essential knowledge the company needs to capitalize (Durstewitz, 1994).(Grundstein, 1995; Grundstein and Barthès, 1996) distinguish on the one hand,know-how (abilityto design, build, sell and support products and services) and on the other hand,individual and col-lective skills (ability to act, adapt and evolve). In a company, there are tangible elements (data,procedures, plans, models, algorithms, documents of analysis and synthesis) and intangible ele-ments (abilities, professional knacks, private knowledge, knowledge of company history and ofdecisional contexts...). Therefore, in a capitalization operation, tangible elements can be takeninto account through KM (technical data management, document management, configurationmanagement), while intangible elements require know-how formalization (acquisition and repre-sentation of know-how and reasoning on such know-how). Know-how, technical facts, productrequirements, design rationale, experience or expertise are examples of knowledge types usefulfor corporate memory (Durstewitz, 1994).

More specifically in the framework of manufacturing industry, several categories of industrialknowledge useful for design activity are proposed in (Bourne, 1997) (see below).1.4

Typologies of Corporate Memories

The memory of an enterprise includes not only a «technical memory» obtained by capitalizationof its employees' know-how but also an «organizational memory» (or «managerial memory»)related to the past and present organizational structures of the enterprise (human resources, mana-gement, etc.) and «project memories» for capitalizing lessons and experience from given projects(Pomian, 1996). (Tourtier, 1995) distinguishes: (a)profession memory, composed of the referen-tial, documents, tools, methods used in a given profession, (b)company memory related to organi-zation, activities, products, participants (e.g. customers, suppliers, sub-contractors), (c)individualmemory characterized by status, competencies, know-how, activities of a given member of theenterprise,project memory comprising the project definition, activities, history and results.(Grundstein and Barthès, 1996) distinguishcompany technical knowledge (i.e. used everydayinside the company, its business units, departments, subsidiaries by the employees for performingtheir daily job) fromstrategic corporate knowledge used by the company managers.

3

1.5Plan of the Paper

The paper will successively analyse problems and solutions linked to detection of needs, cons-truction of the CM, its diffusion, use, evaluation and maintenance. Then we will give severalexamples of dedicated methods and we will summarize the lessons of this study.

2

2.1

KNOWLEDGE MANAGEMENT

Detection of Needs of Corporate Memory

As successful information system development in general, successful CM development must be«underpinned by a clear focus on the situations of use and the needs of users» (Thomas, 1996),i.e. on the human issues of the development. The history of systems development «shows repeate-dly that it is the human issues which «make or break» new methods and tools at work» (Buckin-gham Shum, 1997). So detecting the «right» users' needs, or the «right» CM needed, is the firsttask of the CM designers.

2.1.1Problems

Detecting the «right» needs is not a simple task. CM designers have to learn as much as possibleabout who users are, which tasks they have to perform, in which situations, which knowledgetypes they need to memorize and retrieve (for achieving the tasks), which tools they use, etc. Sodoing, CM designers have to face with problems about users, tasks, situations, etc. Examples ofsuch problems are:

•Users' types:Who are the «right» users to consider? How to take account of the multi-plicity of CM users? Is it worth considering every potential user of the CM? Concerningthe first question, for example, managers of the LJC corporation (a French joint factory)claimed that the customers are important to consider, because they «have the entireknowledge of the product in situation» (Guérin and Mahé, 1997).

•Users' characteristics and behaviours:Which are the «right» users' characteristics andbehaviours to consider? How to «take account of the users' multiple and probably incom-mensurate perspectives» (Kurland and Barber, 1995)? Can we ignore such «side» users'behaviours as «trusting» (Jones and Marsh, 1997)? Which meaningful knowledge storingand knowledge retrieving activities do users perform to achieve their tasks?

•Tasks:Which are the «right» tasks or goals to consider? For example, Simone (1996)identified the following goals of collective memory in the context of dynamic complexsituations: (a) innovating; (b) increasing cooperation; (c) managing turn-over; (d)handling exceptions; (e) dealing with critical situations.

•Situations:Which are the «right» situations, or contexts, to consider? For example,dynamic complex situations (e.g., emergency management, traffic control, rescueservices, industrial plant control) will imply CM requirements different from lessdynamic situations (cf. Wærn 1996).

•Knowledge:Which is the «right» knowledge to consider? Where to get it? What can wedo if the source users (those who have the «right» knowledge) have been transferred, or

4

have resigned, dismissed, or retired (Guérin and Mahé, 1997)?

•Errors:Which are the important CM errors to consider? How to handle them? Forexample, Loftus (1997) reported very interesting studies about false memories showingthat «when people who witness an event are later exposed to new and misleading infor-mation about it, their recollections often become distorted».

CM developers have to face not only with such «first-order» problems (i.e., problems concerningusers directly), but also with «second-order» problems (i.e., problems that directly concern desi-gners). How these «second-order» problems are faced with may have great implications on theneeds detection task. Examples of such problems are:

•CM project ambition:Is the project realistic? A major obstacle to the project achievementis that developers often want «too much, too soon» (Knapp, 1997).

•CM design perspective:Is the goal to create a brand new CM (design), or improving anexisting one (redesign)?

•CM underlying representation:Must CM be considered as an object or as a process (cf.Bannon and Kuutti, 1996)?

•Productivity paradox:How to cope with the productivity paradox, «whereby the availa-bility of more and more information has actually resulted in reducing the production ofthe users» (Sorensen et al., 1997)?

•Context paradox:How to cope with the context paradox, i.e. «the possibility that morecontext will be needed to interpret whatever contextual information has already beenprovided» (Buckingham Shum, 1997)?

2.1.2Solutions

Here are some of the solutions currently adopted to detect CM needs.

2.1.2.1Underlying Approach: «Stakeholder-Centered Design»

The approach to needs detection cannot be disconnected from the approach to the overall develo-pment of the CM, or underlying approach. The main underlying approach is the so-called User-Centered Design (UCD), or Human-Centered Design (HCD), approach. The reason for using aUC[H]D approach is «to ensure that the memory is defined in terms of users' needs» (Durstewitz,1994). The related UC[H]D methods «cover requirements determination, design and implementa-tion, and are concerned with the social as well as technical issues in new system development [...].The philosophy underpinning this approach is that effective systems are created by a partnershipbetween developers and the users and/or stakeholders in the organisation which is to operate thenew system» (Eason and Olphert, 1996). The term«stakeholders» is worth discussing here. Thisterm refers to «any individual within the community where the system may be implemented whohas an interest or «stake» which may be affected by the system» (Eason and Olphert, 1996); itrefers to «anyone who stands to gain from it [the system], and anyone who stands to lose»(Macaulay, 1996). Stakeholders include «potential users but are not restricted to them; other sta-keholders may be purchasers, customers, maintainers, etc. » (Eason and Olphert, 1996). The cur-

5

rent trend among CM developers is to consider stakeholders rather than users (strictly speaking).So CM design/development could be called Stakeholder-Centered Design/Development. As(Eason and Olphert, 1996) claimed: «Systems development should be a partnership in whichdevelopers contribute an understanding of the technical opportunities and the methods of design,and the stakeholders contribute their expertise about the domain of application and existing orga-nisational practices and have a right to judge what is in their best interests as the potential ownersof the future that is being constructed.»

2.1.2.2Approaches to Requirements Analysis

Approaches to needs detection can be appropriately described in terms of requirements analysis,because (1) getting at the users' needs is the aim of requirements analysis (Thomas, 1996), and (2)research on CM and KM often refers to requirements analysis (e.g., Kühn and Abecker, 1997).«The earlier designers of systems understand the needs and problems of their users, and [...] thebetter they understand them then the more able they will be to develop systems which meet users'needs», according to (Macaulay, 1996), that describes four types of approaches to CSCW require-ments analysis (cf. a great amount of CSCW work is done in the context of CM (Wærn, 1996)):Traditional approaches. Traditional approaches are approaches such as the structured analysisapproach, or the object-oriented approach (cf. OO Analysis). In such approaches users have a pas-sive role; they are considered as the sources of information and the reviewers of models develo-ped, and the systems analyst is considered as responsible for eliciting requirements from users.Participation. In the Participation approach, «users are expected to contribute», by assisting inanalysing their problems at work, complete job satisfaction questionnaires, etc. Participation isused «in situations in which initiators of projects do not have all the information needed to designthe change, and where users have considerable power to resist».

Design Team. The formation of a design team is often recommended «to smooth the transitionfrom requirements to design». In the design team, the roles of the technical experts and the custo-mers are clearly identified. Technical experts «contribute their skills to the creation of a system»,and customers «are concerned with the world they will have to inhabit after the change caused bythe system».

Group Sessions. In the Group Sessions approach, people «jointly design systems in facilitatedgroup sessions». Macaulay's cooperative requirements capture (Macaulay, 1996) is a stakeholder-centered approach consisting of the following steps: (1) identify the problem; (2) formulate theteam; (3) group session 1: explore the user environment; (4) validate with users; (5) group session2: identify the scope of the proposed system; (6) validate with stakeholders. Each group sessionhas a number of steps; for example, session 1 includes: (a) the business case, (b) workgroups, (c)users, (d) tasks, (e) objects, (f) interactions, (g) consolidation. Each step includes an introduction,brainstorming, prioritisation and generation of agreed descriptions using checklists and proformaswhich deal with user related issues.

6

It is important to notice that requirements analysis is strongly related to evaluation: if for require-ments analysis the aim is «to get at users' needs», for evaluation the aim is «to tune the system tomake sure that it really does meet those needs» (Thomas, 1996).

2.1.2.3Methods: Classics

Literature Review. Analysing the literature on CM is one of the classical methods used to detectCM needs. For example, from the Macintosh's (1997) work on knowledge asset management,Kühn and Abecker (1997) elicited the following «major impediments to more productivity inknowledge-based work process»: (a) Highly-paid workers spend much of their time looking forneeded information; (b) Essential know-how is available only in the heads of a few employees; (c)Valuable information is buried in piles of documents and data; (d) Costly errors are repeated dueto disregard of previous experiences; (e) Delays and suboptimal product quality result from insuf-ficient flow of information. These impedements can be considered as introducers to requirements.Interviews/Discussions. Performing interviews or discussions is another classical method usedfor identifying CM needs. For example, Kühn and Abecker (1997) had interviews with prospec-tive users and discussions with IT personnel and managers to get requirements. They suggest cru-cial requirements for the success of a CM information system project in an industrial practice: (a)Collection and systematic organization of information from various sources; (b) Integration intoexisting work environment; (c) Minimization of up-front knowledge engineering; (d) Active pre-sentation of relevant information; (e) Exploiting user feedback for maintenance and evolution.Observations/Experiments. Observing real CM practices or conducting experiments aboutthem, are a third classical method used to detect CM needs. For example, observing the DesignRationale activity of a real industrial project conducted in a design office of Aerospatiale, theFrench aerospace company, Karsenty (1996) showed that designers having to reuse a past solutionelaborated by others, often asked themselves: «Why did they do so and not else?» If they had noanswer to this question, experienced designers often considered the alternative solution they spon-taneously found as better than the past one (even if it the later revealed itself worse). Less expe-rienced designers often selected the past solution. These results suggest requirements such as: aCM for Aerospatiale designers should contain justification or argumentation knowledge; thisknowledge must be «past-solution oriented» for experienced designers, and «present-solutionoriented» for less experienced designers.

2.1.2.4Dedicated Methods and Approaches: Some Trends

Lead User Methodology. The «lead user methodology» (Urban and von Hippel, 1988) prescribesto perform needs detection with «lead users». Lead users are «users whose present strong needswill become general in a marketplace months or years in the future».

Advisibility Analysis. The CORPUS project (Grundstein and Barthès, 1996) offers a process-centered and problem-oriented approach called Advisibility Analysis for knowledge capitaliza-tion. The purpose is to help to determine the nature and field of crucial knowledge that needs to becapitalized, the company members who have this knowledge, in which form, the members whouse this knowledge, when and how, and the risks in case no capitalization operation is performed.

7

The main steps of this approach are: (1) Determine sensitive processes essential for the companyfunctioning; (2) Distinguish determining problems that fragilize critical activities (i.e. activitiescontributing to sensitive processes); (3) Determine crucial knowledge necessary to solve determi-ning problems.

Enterprise Models. Some research focus on enterprise analysis and modelling (Fox, 1993) [http://www.aiai.ed.ac.uk/~entprise/enterprise/] and can be useful during a CM construction. For exam-ple, the evolution of the enterprise through time, its experience acquired from past projects areelements interesting to take into account. An enterprise ontology, defining concepts relevant fordescription of an enterprise, is proposed in (Uschold, King, Moralee, and Zorgios, 1996). Such anontology can be used as support for exchange of information and knowledge in the enterprise(Fraser, 1994). Organizational structure, processes, strategies, resources, goals, constraints andenvironment of the enterprise can thus be modelled. Intra-enterprise modelling and inter-enterpri-ses modelling can be distinguished. (Beauchène, Mahé and Rieu, 1996) models an enterpriseorganization, using a model stemming from quality management and focusing on «customer-sup-plier» relationships between the enterprise members. The interest of exploiting an enterprisemodel is to determine the weak points of the enterprise, that could possibly be improved by aknowledge capitalization operation.

The MNEMOS EUREKA project (see http://www.delab.sintef.no/MNEMOS/dir.html) aimed «todevelop a new generation of information systems to increase the competitivity of the enterprisethrough a better circulation of the corporate knowledge, a more efficient management and supportto the human creativity processes». This project proposed an enterprise model based on eightdimensions: document, programme, budget, contacts, organization, material, calendar, results(Feray and Villefranche, 1997).

Cognitive Models. Theoretical models of workers' cognitive functioning and of knowledge usedin work situations may be useful for needs detection purposes. (Bollon, 1997) showed the interestof these models and especially the methodological precautions they induce during field observa-tions conducted to capitalize knowledge (see also Poitou, 1997.)

Anthropotechnology. Anthropotechnology (Wisner, 1997) refers to the transfer of organisationalsystems and technologies in countries having different cultures. This methodology can be appliedto design within the same country or the same organisation, in which different-culture subgroupscan be identified. From the anthropology viewpoint, culture-related requirements need to be iden-tified for a successful transfer.

Knowledge Networking. From the point of view of expertise sharing between CM developers, aproject which anticipates what would happen in the future of CM development practice isCERES-GKN [http://www.cerc.wvu.edu/ceres/CERESGKN_brochure.html]. The goal of thisproject is to construct «a global knowledge network to enable environmentally sound product andprocess development». CERES-GKN «will identify consumer and producer requirements andneeds for an environment-oriented infrastructure and product and process application». CERES-GKN «will develop a global network of knowledge bases (both proprietary and public domain)containing a variety of knowledge -- such as best practices, case studies, and expert advisory sys-

8

tems -- useful for designing products and processes that are at once environmentally sound, tech-nologically feasible, and economically justifiable».

2.1.3Conclusion

The phase of needs detection may help to determine the type of CM needed (e.g. project memory,profession memory, organizational memory, individual memory), the potential users of the CM,and the possible modes of exploitation useful and adapted to their work environment.2.2

Construction of the Corporate Memory

As emphasized during KAW'96 track on «Corporate Memory and Enterprise Modelling», a cor-porate memory is of course different from a knowledge-based system. The techniques adopted tobuild a CM depend on the available sources: human specialists, existing paper-based or electronicdocuments such as reports or technical documentation, E-mails, existing databases, case libraries,dictionaries, CAD drawings... They also depend on the nature of the needed CM according to theintended users: it may consist of paper-based documents making explicit the enterprise adequatemembers' knowledge, that had never been yet elicited and modelled (Dieng, Giboin, Amergé,Corby, Després, Alpay, Labidi, Lapalut, 1996). It may also be a computational memory materiali-zed through an intelligent documentary system, a knowledge base, a case-based system, a Web-based system or a multi-agent system. We note that even though paper-based or electronic docu-ments can themselves represent a CM they are often considered as a first step in the implementa-tion of the CM (Simon,1996).

In the next sections, we describe different approaches for the construction of a CM.

2.2.1Non computational Corporate Memory

Anon computational memory is made of paper-based documents on knowledge that had neverbeen elicited previously. The construction of such a memory may be guided by two different aims:(a) to elaborate synthesis documents on knowledge that is not explicit in reports or technical docu-mentation, and is more related to the know-how of the experts of the enterprise, (b) to improveenterprise production through expert propositions on their tasks in a design process.

In the first aim, the memory is composed of knowledge described in existing documents and inter-views of experts, or elaborated from observations of experts’ activities. The KADE-TECH Com-pany proposes a method called CYGMA (Bourne, 1997) to produce different documents thatcontain memory about a profession (see below). (Simon, 1996) considers that this kind ofmemory provides «a global view of the knowledge of the firm», and «allows experts from diffe-rent sites to describe their knowledge in the same format in order to be able, afterwards, to com-pare them more easily». But in (Simon, 1996), this elaboration of synthesis documents is a firststep in the construction of the computational CM that it helps to implement: it enables homoge-neization of know-how in different sites of an enterprise distributed geographically.

In the second aim, the firm RENAULT proposes MEREX approach (Corbel, 1997). Thisapproach, guided by the Quality approach, is based on positive and negative experience return onprevious projects. The memory is constituted by forms, where an expert can describe a solution ora decision in a task of design process. Those forms are validated by a system of check-list and sto-red in a form management system. They are used in the product specification phase, before the

9

artefact design.

Remark: Notice that often such paper-based documents are put later in an electronic form, but wemake a difference between simple electronic documents and an actual documentary system.

2.2.2Document-based Corporate Memory

A document-based CM relies on the principle that all existing documents of the firm can consti-tute the CM. But those documents are not well-indexed or they constitute a personal bibliographyfor each expert of the firm. So the construction of such a CM begins by indexing all reports, syn-thesis documents or references used by the different experts. It requires an interface to managedocuments (addition of documents, retrieval of documents...). (Poitou, 1995) considers that: «agood documentation system is very likely the least expensive and the most feasible solution toknowledge management» and prefers a computer assistant to documentation (i.e. to writing orreading) rather than knowledge representation: according to him, a document is already a repre-sentation of knowledge. So the main need is assistance in preparing, storing, retrieving and pro-cessing documents. The notion of corporate knowledge collective management system (Poitou,1997) answers well to this need: e.g. SG2C proposed by Poitou and DIADEME proposed byElectricité de France (Ballay and Poitou, 1996; Ballay, 1997).

2.2.3Knowledge-based Corporate Memory

Knowledge engineering is naturally useful for building a CM based on elicitation and explicitmodelling of knowledge from experts or even for a formal representation of knowledge under-lying a document. Therefore several researchers that have been working on expert systems foryears evolved towards CM building where they could exploit their past experiences. However, thegoal of a CM building is less ambitious than an expert system: instead of aiming at an automaticsolution for a task (with automatic reasoning capabilities), a CM rather needs to be an assistant tothe user, supplying him/her with relevant corporate information but leaving him/her the responsi-bility of a contextual interpretation and evaluation of this information (Kuhn and Abecker, 1997).(Kuhn and Abecker, 1997) notices that «in contrast to expert systems, the goal of a CM is not thesupport of a particular task, but the better exploitation of the essential corporate resource:knowledge» but, however, cites some knowledge-based CM implemented through an expert sys-tem (e.g. KONUS system aimed at support to crankshaft design).

Knowledge engineering methods such as COMMET and CommonKADS can be useful in theconstruction of a CM, because they allow to analyse and represent an activity on the knowledgelevel. (Steels,1993) notices that the organization of a production is more and more horizontal, i.e.the production is organized through activities gathering experts stemming from different depart-ments. So the CM of such an enterprise can be based on activity description through three pers-pectives: task, method and information and can thus be realized using KREST. By the same way,even though CommonKADS was not primarily dedicated to CM building, some models offeredby CommonKADS (organization, task, agent, communication and expertise models) give an inte-resting basis for knowledge-based CM (Kingston, 1994; VanderSpek, 1994; Corby and Dieng,1997).

10

2.2.4Case-based Corporate Memories

The exploitation of another AI technique, case-based reasoning, can also be very useful (Simonand Grandbastien, 1995; Simon, 1996). Indeed each firm has a collection of past experiences (suc-cesses or failures) that can be represented explicitly in a same representation formalism allowingto compare them. The use of a case base for representing the CM is dedicated for the followingaims: (1) avoid the scattering of the expertise by concentrating knowledge of all experts in dedi-cated cases, (2) allow a continuous evolution of the CM thanks to the progressive addition of newcases.

Case-based reasoning allows to reason from experiences and cases already encountered, in orderto solve new problems: e.g. for maintenance of a complex equipment, the collective memory ofpast incidents can be useful for taking a decision in case of a new breakdown. The retrieval of asimilar past case can be used to suggest a solution to a new problem to be solved (this solution canbe reused or adapted if needs be). Improving representation of the cases, organization andindexing of the case base is important for enhancing efficiency of case retrieval.

In (Simon, 1996 ; Simon, 1997), the author describes an example in metallurgy, where the aimwas to capitalize knowledge and know-how about descriptions of production of produced steelsand metallurgical defects encountered during these productions. The purpose of the CM was toexploit past successes and failures in order to minimize error risks in design of new steels. Themethod consisted of: (1) creating synthesis documents common to all sites and respecting anhomogeneous format, (2) proposing models to implement a CM based on such synthesis docu-ments, (3) providing capitalization processes allowing to use the CM for defects detectionpurpose.

2.2.5Construction of a Distributed Corporate Memory

A distributed CM is interesting for supporting collaboration and knowledge sharing between seve-ral groups of people in an organization or in several collaborating organizations, such groupsbeing possibly dispersed geographically. A distributed memory is essential for virtual enterprisesmade of distributed organizations and teams of people that meet and work together online. Gene-rally, for such virtual enterprises, this distributed memory naturally relies on the exploitation ofthe Internet and of the Web (O'Leary, 1997). For example, the GNOSIS project on intelligentmanufacturing (Gaines, Norrie, Lapsley and Shaw, 1996) involves several enterprises distributedthrough several continents. Coordination of this project and management of distributedknowledge among the participants is performed through the Web. The tools developped in theproject are used for keeping a memory of the project.

A distributed CM can be made of distributed, heterogeneous knowledge bases or of distributed,heterogeneous case bases, or of a multi-agent system. For example, in the MEMOLAB project,the CM of a research laboratory is implemented through a multi-agent system (with agents suchas a bibliographic agent, a notebook agent, a «tips and tricks» agent and a proxy agent) (Vanden-berghe and de Azevedo, 1995). The implementation of a distributed memory can also rely onboth distributed case libraries and artificial agents responsible for information retrieval amongsuch libraries (Nagendra Prasad and Plaza, 1996).

The construction of a distributed CM may often involve several experts. A protocol for collectiveknowledge elicitation is proposed in (Dieng, Giboin, Amergé, Corby, Després, Alpay, Labidi,

11

Lapalut, 1996). Problems of consistency of the obtained CM elements, of cohabitation of severalviewpoints must be solved: a protocol for cooperative creation of a consensual CM is thus offe-red in (Euzenat, 1996). In the particular case of a distributed CM relying on the reuse of ontolo-gies, research on the collaborative creation of ontologies via ontology servers such as Ontolingua(Farquhar, Fikes and Rice, 1996), APECKS (Tennison and Shadboldt, 1998) or WebOnto(Domingue, 1998) can be exploited.

2.2.6Combination of Several Techniques

In some cases, both informal knowledge (such as documents) and formal knowledge (such asknowledge explicitly represented in a knowledge base) are needed. Therefore research on themanagement of links between document and knowledge base can be exploited (Martin andAlpay, 1996; Euzenat, 1996). By the same way, research on the semi-automatic extraction ofknowledge (for example, terminological knowledge, etc.) from documents thanks to natural lan-guage analysis can be useful (Trigano, 1994). (Kühn and Abecker, 1997) proposes an interestingCM architecture where the CM can be composed of different sorts of memories: documents,knowledge bases, etc.2.3

Diffusion and Use of the Corporate Memory

2.3.1Possible Modes of Diffusion

Adequate elements of the CM must be distributed to the adequate members of the enterprise: thisdistribution may be passive or active, as either the user can search by himself needed informationwhere it is available, or knowledge distribution can be systematically decided and taken in chargeby an adequate person or service of the enterprise. When the company workers are too busy tolook for relevant corporate information, a passive distribution is insufficient: (Kuhn and Abecker,1997) recommends an active distribution (e.g. a regular recall of the existence of relevant informa-tion). (Van Heijst, Van der Spek, and Kruizinga, 1996) distinguishes several cases according to thekind of collection and of diffusion of the CM :

•Knowledge attic:both collection and diffusion are passive. It corresponds to the case of aCM used as an archive which can be consulted when needed.

•Knowledge sponge:the collection is active but the diffusion is passive.

•Knowledge publisher:the collection is passive but the distribution is active, as the CMelements are forwarded to people for whom they will be relevant.

•Knowledge pump:both collection and diffusion are active. For example, in ICAREproject (Bologna and Gameiro Pais, 1997), in each department of the company, a«knowledge watcher» is responsible for planning the knowledge element collection fromhis/her department and for inciting the members of this department to consult the CM.

2.3.2Diffusion via Intranet / Internet

Individuals and organizations can take advantage of the remarkable possibilities of access to data,to information and to knowledge provided by Internet. Knowledge diffusion can for exampleexploit the possible access to Internet or to an Intranet inside the enterprise. However, sometimes

12

some reticences are expressed by the managers of some enterprises w.r.t. Internet and the Web,due to potential problems such as confidentiality, security, reliability of accessed information, riskof information excess that may disturb the employees in their work. But security problems arestudied actively by researchers, as they are a significant condition for success of Internet-basedapplications such as electronic commerce.

Diffusion can rely on a knowledge server on the Web or on publication on the Web (Euzenat,1996; Corby and Dieng, 1997). Different kinds of elements can be accessed through Internet/Intranet: documents (classic electronic documents, HTML documents...), databases, ontologies,knowledge bases, case bases, articles of electronic journals, etc. Therefore several kinds ofknowledge servers can be thought out: document servers, ontology servers, knowledge base ser-vers, database servers, journal servers or digital libraries. The main problems to be solved areretrieval of elements of the CM in answer to a request and adaptation of the answer to the user.Research on user profiling can thus be useful in this purpose (Sorensen, O'Riordan and O'Riordan,1997).

Let us notice that a CM may not be restricted to the enterprise, but could include information andknowledge stemming from the external world but important for the enterprise work (cf. the so-cal-led «economic intelligence»). Therefore, the retrieval and integration of such information expli-citly put on the Web by other companies working in the same area may be interesting.

2.3.2.1Example of Diffusion via Internet/Intranet

Let us detail an example of exploitation of Internet/Intranet. In our team, we have developped acomponent, called WebCokace, that enables to distribute expertise on the Internet (Corby andDieng, 1997). The expertise is modelled in the CommonKADS framework (Breuker and Van deVelde, 1994) with the CML formalism (Schreiber, Wielinga, Akkermans, van de Velde, andAnjewierden, 1994). WebCokace relies on the hypothesis that CommonKADS may be useful forbuilding knowledge-based corporate memories.

WebCokace takes advantage of the Web technology to interface an expertise model developmentenvironment with an HTTP server. The expertise model environment functions in a server modeand is connected to an HTTP server (that acts here as a client of the knowledge server) by meansof a CGI interface. Modelled knowledge is then available on the Net.

In order to facilitate user interaction with the system, we have developped a search engine, aquery language and an interpreter for this language. Users can emit queries to the knowledge ser-ver and get CommonKADS objects in response to the queries. CommonKADS objects are pretty-printed with HTML hypertext links to related objects in such a way that hypertext navigation ispossible in expertise models. For example, a concept references its super types, a task its subtasks.The system generates interactive graphic views on the expertise. It is possible to visualize conceptand task hierarchies, domain models, etc. Clicking on a node of a hierarchy leads to the corres-ponding object definition. So the end-user may rely on the graphics instead of CML text.

The system also manages references between expertise models and electronic documents bymeans of HTML hypertext links and URL. A CommonKADS model can be annotated with refe-rences to source documents (e.g. technical documentation, articles, etc.), and conversely, a docu-ment can be annotated with references to expertise models. The links are activated once loaded ina Web browser and it is then possible to navigate between models and documents in a hypertextway.

13

Using WebCokace, we have developped a generic library for conflict solving in concurrent engi-neering, an oncology server and we have implemented parts of the CommonKADS modelinggeneric library.

WebCokace is implemented on the Centaur programming environment generator, developed inthe Croap project at the INRIA. Thanks to the underlying generic technology (i.e. Centaur), Web-Cokace can be used as aprogram server for any programming language that is implemented inCentaur. Within Centaur, programs are internally manipulated as abstract syntax trees (AST).AST support abstract computations on programs that enable to answer to queries. A program ser-ver can be useful in companies having libraries of programs to be included in their CM.

2.3.3Information Retrieval

The CM is supposed to be used by adequate members of the enterprise: in all cases (documentarysystem, knowledge base, case-based system, Web-based system, etc), we must notice the impor-tance of information search, if possible adapted to the users' needs, their activities and their workenvironment. The problems to be tackled are: How can the user express his/her requests? How toimprove hypertext navigation by the user? How to retrieve elements of the CM in answer to arequest? Is full-text search sufficient? How to index the documents to retrieve? What additionalinformation (such as enterprise models, knowledge models, user models) could help to filter theinformation to be retrieved? Are inference capabilities needed in this purpose? Research on onto-logy servers such as Ontolingua (Farquhar, Fikes and Rice, 1996), APECKS (Tennison and Shad-boldt, 1998) or WebOnto (Domingue, 1998) could also be exploited, since a part of the CM canrely on an ontology. A CM infrastructure relying on techniques of information search on the Inter-net is proposed in (Huynh, Popkin and Stecker, 1994).2.4

Evaluation and Evolution of the Corporate Memory

2.4.1Evaluation of the Corporate Memory

As noticed in (Ermine, 1996), operational projects of CM are necessarily consuming and expen-sive. Therefore an evaluation of such projects is important, from several viewpoints: economico-financial, socio-organizational and technical.

From aneconomico-financial viewpoint, one aim of the CM is to improve the enterprise competi-tiveness. As noticed in (Durstewitz, 1994), it can be measured by a gain between the success ofthe enterprise products or services, and its production (and maintenance) costs. There must be anevaluation of the gain obtained thanks to the introduction of a CM, generally aimed at enhancingproductivity. Return on investment is important for justifying the interest of building a CM, fromthe viewpoint of the managers. But methods or tools are needed to assess the actual improvementdue to the introduction of the CM: it may be an improvement in safety - cf. avoidance of pasterrors -, in quality and in performance.

From asocio-organizational viewpoint, the CM can aim at improving employees' work organiza-tion (thanks to information circulation improvement, etc.) and employees' satisfaction in theirwork. But the criteria for such an evaluation are often qualitative and hardly quantitative: they canrely on classical criteria used for evaluating user-centered tools such as easiness of use, easinessof information retrieval, adequation of retrieved information, confidence in such information, usa-

14

bility for the user's activity, etc. As noticed in (Kuhn and Abecker, 1997), users’ feedback shouldbe exploited for detecting possible deficiencies of the CM and suggest improvements of the CM.From atechnical viewpoint, the transfer of know-how inside the enterprise seems to be an evidentbenefit. But an effective transfer depends on an effective use of the CM and on its adaptation tosuch a knowledge transfer.

There may be some bias in the use of the CM. The introduction of a CM can imply changes inindividual and collective work in the enterprise. Some reorganizations prescribed by the managersmay not be accepted by the employees. For example, an official procedure for storing lessons orexperiences linked to a given project may be prescribed by the company managers but not respec-ted for reasons such as lack of time, lack of motivation, etc. Moreover, a CM may be used othe-rwise than planned. We found very few publications analyzing reactions of CM users: forexample, in (Ballay and Poitou, 1996), a survey of satisfaction of DIADEME users is presented. Itrelied on a questionnaire on their use of automatic bibliography and hypertext links, their expe-rience and satisfaction of the databases, their experience and satisfaction with the full-text docu-ment retrieval TOPIC included in DIADEME, their satisfaction with the workstation. The lessonof this survey was that even though DIADEME was aimed at being a collective knowledge mana-gement system, its users rather exploited the system as a set of different specific tools. In (Kuhnand Abecker, 1997), three case studies are analyzed: KONUS for crankshaft design, RITA forQuality Assurance for Vehicle Components and PS-Advisor for bid preparation for oil productionsystem. The authors noticed that all three systems failed to go beyond prototype stadium and beintegrated in the company daily operational work. The reasons of such failures were: «costs ofcustomer-tailored solutions with unpredictable return of investment, insufficient experiences withCM applications, poor integration into the conventional Information Technology landscape». As alesson learnt from these case studies, they suggested crucial requirements for a CM (see above),they proposed a general CM architecture and a kind of methodological guide for development of aCM, insisting on requirement analysis, human factors, cost-benefit analysis, knowledge evolutionand technical realization.

In conclusion, we must distinguish evaluation by users (with criteria based on users' satisfaction)and strategic evaluation by managers (with criteria based on return on investment). At present,there are too few effective operational CM, and companies need to stand back for evaluating themprecisely.

2.4.2Maintenance and Evolution of the Corporate Memory

For maintenance and evolution of the CM, it is necessary to take into account the results of theevaluation of what already exists. Problems linked to addition of new knowledge, removal ormodification of obsolete knowledge, coherence problems underlying a cooperative extension ofthe CM, must be tackled. Some of such problems were already relevant during the construction ofthe CM. Likewise, both organizational problems and technical problems underly the possible evo-lution of the CM. In the construction as in the evolution of the CM, some problems may stemfrom conflicts between persons, reticences, lack of motivation, lack of time.

The techniques used to maintain and make evolve the CM also depend on the kind of CM: accor-ding to the case, addition, removal or modification will concern elements of a knowledge base orcases in a case base or (elements of) documents in a document base or agents in a multi-agent sys-tem. The CM evolution also depends on whether the collection (resp. diffusion) of CM elements

15

is passive or active (Van Heijst, Van der Spek, and Kruizinga, 1996). Evolution of the CMdepends on both the CM builders/maintainers and the CM users.

According to (Kuhn and Abecker, 1997), knowledge evolution should be «a continuous activityperformed by a CM administrator in close cooperation with the users who can make improvement/ update suggestions tightly integrated into their work process». This solution corresponds to anactive collection and diffusion, as for instance in the ICARE project (Bologna and Gameiro Pais,1997). In some cases, a given service or a given person of the enterprise is responsible for themaintenance/evolution of the CM. In other cases, any employee may make evolve the CM, whilerespecting some constraints.

3EXAMPLES OF DEDICATED METHODS

This section will give few examples of methods dedicated to the building of a CM. The purpose ofthis description is to show the principles guiding some CM-dedicated methods (in comparison toknowledge engineering methods such as KREST or CommonKADS).3.1

Method CYGMA (KADE-Tech)

CYGMA (CYcle de vie et Gestion des Métiers et des Applications) is a method allowing theconstruction of a profession memory in a manufacturing industry (Bourne, 1997). It defines sixcategories of industrial knowledge for design activity:

•singular knowledge:positive and negative, relevant or out of bound experiences;•terminological knowledge:alphabetical list of terms used in the profession domain;

•structural knowledge:it contains the ontological knowledge, and a factual knowledgebase comprising the initial data of the design problem to be solved and the initial goalsdescribing the design problem solution to be found;

•behavioural knowledge:dynamic elements of profession knowledge;

•strategic knowledge:knowledge allowing an optimized use of structural and behaviouralknowledge;

•operating knowledge:knowledge describing the problem solving process as a chaining ofoperating activities based on structural, behavioural and strategic knowledge.

The results of the method application consists of four different documents:profession glossarygathering singular and terminological knowledge,semantic catalogue describing structuralknowledge,rule notebook comprising behavioural knowledge,operating manual made of strate-gic and operating knowledge. These documents can then be exploited by the enterprise as a wayof communication with subcontractors. The method has already been applied to different profes-sions in different firms: blacksmith profession for Rolls-Royce, turner profession for Eurocopter,automatician profession for Fiat and steel manufacturer profession for Aérospatiale.3.2

Method REX (CEA)

REX method (Malvache and Prieur, 1993) relies on the following steps: (1) analyse needs andidentify sources of experience, (2) constructelementary pieces of experience from documents,

16

databases or interviews, (3) build up a computer representation of the knowledge domain, (4) ins-tall a software package on the user’s workstation: this package includes a multimedia interfaceand aretrieval engine that produces information files on the basis of questions in natural lan-guage.3.3

Method MKSM (CEA)

MKSM (Method for Knowledge System Management) (Ermine, 1996; Ermine, Chaillot, Bigeon,Charreton, Malavieille, 1996) aims at reducing complexity of knowledge system management,using different models at different grain levels. It is a systemic-based decision support method. Itrelies on the hypothesis that the knowledge assets of an organization can be considered as acom-plex system. Modelling such a complex system relies on several viewpoints: syntax, semantic andpragmatic, each viewpoint being itself modelled through three viewpoints: structure, function andevolution. The three components of a knowledge sytem areinformation (requiring data proces-sing),signification (requiring task modelling) andcontext (requiring activity modelling). Themethod offers five modelling phases: knowledge system modelling, domain modelling, activitymodelling, concept modelling, task modelling.3.4

Comparison of the Methods

CYGMA is dedicated to profession memory, in the framework of a design task, while REX andMKSM do not focus on a kind of CM and do not restrict to a kind of task. REX relies on the buil-ding of pieces of experience, stemming from several kinds of sources (human, documents, databa-ses); such pieces can be retrieved in answer to natural language request. MKSM takes inspirationof complex system theory for offering a theoretical analysis of an organization knowledge, consi-dered as a complex system. MKSM proposed modelling phases are close to CommonKADSnotions. All three methods were applied to several industrial applications. Criteria for comparingthem more precisely could be : the complexity level of the method application, the kind of CM itenables to build, the kind of task it restricts to, the number and features of effective applicationsbuilt with them, and evaluation of such applications by their end-users.

4CONCLUSIONS

We presented a preliminary survey that needs to be completed and deepened in order to offer amethodological guide for the choice between the multiple methods and tools proposed. We canhowever propose some conclusions from this survey.

In all the described research, an important aspect is thatan enterprise can be analyzed at severallevels, according to several viewpoints. Most methods focused on some viewpoints and relied onan implicit or explicit model of the enterprise, or at least of the knowledge of the enterprise. Theanalysis of the enterprise needs for a CM can help determine the kind of needed CM. Accordingto the case, it may imply to build an individual memory (cf. an expert retires or is muted, so it isinteresting to make explicit, model and store this expert's know-how in a knowledge base or tostore his experiences in a case base), a project memory (cf. elements of a given project could benecessary for later projects), a managerial memory needed by the company managers for strategicdecisions, etc.

17

As a conclusion, the choice between the different construction techniques is based on severalquestions that an enterprise should answer before building a CM:

•What is the knowledge already existing in the firm?

•What kind of knowledge must contribute to the construction of the CM?:

-knowledge already described in documents as reports or synthesis document on a project?- professional knowledge not already described in documents?•What is the intended use of the CM after its construction: is it...

- a way of communication between geographic distant sites?

- a way of communication between an enterprise and subcontractors?- a way to enhance learning of the enterprise members?

•Is it necessary to model knowledge of some enterprise members or is an intelligentdocumentary system sufficient?

The ability to store and reuse both knowledge, elements of experience and documents is impor-tant, as well as the ability to offer user-tailored information retrieval.References

Ballay, J.-F. (1997). Capitaliser et transmettre les savoir-faire de l'entreprise, Eyrolles.

Ballay, J.-F., and Poitou J. P. (1996). Diadème : a Collective Knowledge Management System. In J. F. Schreinemakers ed,

Knowledge Management: Organization, Competence and Methodology, Proc. of the 4th Int. Symposium on the Managementof Industrial and Corporate Knowledge (ISMICK'96), Rotterdam, the Netherlands, Wurzburg:Ergon Verlag, Advances inKnowledge Management, vol. 1, October 21-22, p. 265-285.

Bannon, L. J. and Kuutti, K. (1996). Shifting perspectives on organizational memory: from storage to active remembering. In Y.

Wærn (1996)

Barthès, J.-P. (1996), ISMICK and Knowledge Management. In J. F. Schreinemakers ed, Knowledge Management: Organization,

Competence and Methodology, Proc. of ISMICK'96, Rotterdam, the Neth., Wurzburg:Ergon Verlag, October 21-22, p. 9-13.Beauchène, D., Mahé, S., Rieu, C. (1996). Enterprise Know-How: Capitalization and Benchmarking with an Enterprise Organiza-tional Model. In J. F. Schreinemakers ed, Knowledge Management: Organization, Competence and Methodology, Proc. ofISMICK'96, Rotterdam, the Netherlands, Wurzburg:Ergon, October 21-22, p. 194-206.

Bollon, T. (1997). Capitalisation des connaissances et conception: éléments de méthodologie. In J.-M. Fouet (Coord.), Connais-sances et savoir-faire en entreprise: information et capitalisation, Paris, Hermès, pp. 130-153.

Bologna , G., and Gameiro Pais, J. (1997). ICARE: an operational knowledge management system. Proc. of the 5th Int. Sympo-sium on the Management of Industrial and Corporate Knowledge (ISMICK'97), Compiègne, France, p. 150-162.

Bourne, C. (1997). Catégorisation et formalisation des connaissances industrielles. In Connaissances et Savoir-faire en entreprise,

ed. Hermès, p 179-197.

Breuker, J. and Van de Velde, W., eds (1994). CommonKADS Library for Expertise Modelling, Reusable Problem Solving Com-ponents. IOS Press, Amsterdam.

Buckingham Shum, S. (1997). Negotiating the construction and reconstruction of organisational memories. Journal of Universal

Computer Science, 3(8), 899-928.

Corbel, J.C (1997). Méthodologie de retour d'experience: démarche MEREX de Renault. In J.-M. Fouet (Coord.), Connaissances

et Savoir-faire en entreprise, ed. Hermes, pp. 93-110.

Corby , O. , and Dieng R. (1996). Cokace: a Centaur-based environment for CommonKADS Conceptual Modelling Language. In

W. Wahlster ed, Proc. of ECAI'96, pp. 418-422, John Wiley & Sons, Ltd. Budapest, Hungary, 1996.

Corby , O. , and Dieng R. (1997). A CommonKADS Expertise Model Web Server. Proc. of the 5th Int. Symposium on the Mana-gement of Industrial and Corporate Knowledge (ISMICK'97), Compiègne, France, p. 97-117. (WebCokace is available at theURL: http://www.inria.fr/acacia/Cokace)

18

Dieng, R., Giboin, A., Amergé, C., Corby, O., Després, S., Alpay, L., Labidi, S., Lapalut, S. (1996). Building of a Corporate

Memory for Traffic Accident Analysis. In Gaines & Musen eds, Proc. of KAW'96, Banff, Canada, November, p. 35.1-35.20.Domingue, J. (1998).Tadzebao and WebOnto: Discussing, Browsing, and Editing Ontologies on the Web, Accepted at the 11th

Workshop on Knowledge Acquisition, Modeling and Management (KAW'98), April 18-23.

Durstewitz, M. (1994). Report on Workshop on Corporate Memory, Toulouse, http://www.delab.sintef.no/MNEMOS/external-info/cm-eurisko.txt

Eason, K. and Olphert, W. (1996). Early evaluation of the organisational implications of CSCW systems., in P. Thomas (Ed.) ,

CSCW Requirements and Evaluation, London: Springer, pp. 75-89.Ermine, J. L. (1996) Les systèmes de connaissances. Hermès.

Ermine, J. L. , Chaillot, M., Bigeon, P., Charreton, B., Malavieille, D. (1996) . MKSM, a method for knowledge management. In J.

F. Schreinemakers ed, Knowledge Management: Organization, Competence and Methodology, Proc. of ISMICK'96, Rotter-dam, the Netherlands, Wurzburg:Ergon, Advances in Knowledge Management, vol. 1, October 21-22, p. 288 - 302.

Euzenat, J. (1996) . Corporate memory through cooperative creation of knowledge bases and hyper-documents. In B. Gaines, M.

Musen eds, Proc of KAW'96, Banff, Canada, November, pp. 36-1 36-18.

Farquhar, A., Fikes, R. and Rice, J. (1996). The Ontolingua Server: a Tool for Collaborative Ontology Construction. In B. Gaines,

M. Musen eds, Proc of KAW'96, Banff, Canada, November, pp. 44-1 – 44-19.

Feray, N. , Villefranche, L. (1997). Modèle des Données et de la Mémoire de l'Intranet MNEMOS, Technical Report, Laboratory

PSI, Rouen.

Fox, M. S. (1993). Issues in Enterprise Modelling. In Systems Engineering in the Service of Humans, Proc. of the IEEE Conf. on

Systems, Man and Cybernetics, p. 86-92, Le Touquet, France, October 17-20. Los Alamitos, CA : IEEE Computer Society.Fraser, J. (1994). Managing Change through Enterprise Models, in Proc. of Expert Systems’94, the 14th Annual Conference of the

British Computer Society Specialist Group on Expert Systems, Cambridge UK, December.

Gaines, B. R., Norrie D. H., Lapsley A. Z., Shaw M. L. G. (1996). Knowledge Management for Distributed Enterprises. In B. Gai-nes, M. Musen eds, Proc of KAW'96, Banff, Canada, November, pp. 37-1 37-18.

Grunstein (1995). La capitalisation des connaissances de l'entreprise, système de production de connaissances. L'entreprise appre-nante et les Sciences de la Complexité. Aix-en-Provence, Mai.

Grunstein, M. and Barthès J.-P. (1996). An Industrial View of the Process of Capitalizing Knowledge. In J. F. Schreinemakers ed,

Knowledge Management: Organization, Competence and Methodology, Proc. of ISMICK'96, Rotterdam, the Netherlands,Wurzburg:Ergon Verlag, Advances in Knowledge Management, vol. 1, October 21-22, p. 258-264.Guérin, C. and Mahé, T. (1997). Entreprises, exercez votre mémoire! Sciences et Techniques, No 784.

Huynh M., Popkin L., Stecker M. (1994). Constructing a Corporate Memory Infrastructure from Internet Discovery Technology.

White Papers, Marble Associates Inc., .

Jones, S., Marsh, S. (1997). Human-Computer-Human Interaction: Trust in CSCW. SIGCHI Bulletin, 29(3), 36-40.

Karsenty, L. (1996). An empirical evaluation of design rationale documents, Electronic Proceedings of CHI'96 [http://

www.acm.org/sigchi/chi96/proceedings/papers/Karsenty/lk_txt.htm]

Kingston, J. K. C. (1994). Modelling Business Processes using the Soft Systems Approach. In J. P. Barthès ed., Proc. of the 2nd

Int. Symposium on the Management of Industrial and Corporate Knowledge (ISMICK'94), Compiègne, October 1994, pp.149-159.

Knapp, E. (1997). Know-how's Not Easy: How to Keep knowledge management from flickering out. ComputerWorld, Leadership

Series, 1/17/97.

Kühn O., Abecker A. (1997). Corporate Memories for Knowledge Management in Industrial Practice: Prospects and Challenges.

Journal of Universal Computer Science, 3(8), 929-954.

Kurland, T. Barber, P. (1995). User requirements from a group perspective: the case of distance learning mediated by computer

conferencing. In Peter Thomas (ed), CSCW Requirements and Evaluation, London: Springer, 57-74.Loftus, E.L. (1997). Creating false memories. Scientific American, September 1997, 70-75.

Macaulay, L. (1996). Cooperation, Requirements Analysis and CSCW, in P. Thomas (Ed.) CSCW Requirements and Evaluation,

pp. 39-55.

Macintosh, A. (1994). Corporate Knowledge Management State-of-the-Art Review. Proc. of ISMICK'94, Compiègne, p. 131-145.Macintosh, A. (1997). Knowledge asset management, AIring, (20), April 1997.

Malvache, P. and Prieur, P. (1993). Mastering corporate experience with the Rex method. In J. P. Barthès ed., Proc. of ISMICK'93,

Compiègne, October, pp. 33-41.

19

Martin, Ph., and Alpay, L. (1996). Conceptual Structures and Structured Document. In Eklund, P. W. , Ellis, G., Mann, G. eds,

Conceptual Structures: Knowledge Representation as Interlingua, ICCS'96, Springer-Verlag, LNAI 1115, pp. 145-159.

Morizet-Mahoudeaux, P. (1994). Basic Research Knowledge Management in Industry, Academia and Administration: A Japanese

Perspective. Proc. of ISMICK'94, Compiègne, France, pp. 3-18.

Nagendra Prasad, M.V. N., Plaza, E. (1996), Corporate Memories as Distributed Case Libraries, Proc. of KAW'96, Banff, Alberta,

Canada, November 9-14, p. 40-1 40-19. Also in http://ksi.cpsc.ucalgary.ca/KAW/KAW96/KAW96Proc.htmlNonaka, I. (1991). The knowledge-creating company. Harvard Business Review. Nov-Dec., pp. 96-104.O'Leary, D. E. (1997). The Internet, Intranets and the AI Renaissance. Computer, 30(1):71-78, Janvier.

Oliveira E. and Shmeil M. A. H. (1995). ARTOR - An Approach towards Artificial Organizations. Proc. of the 3rd Int. Symposium

on the Management of Industrial and Corporate Knowledge (ISMICK'95), Compiègne, France, pp. 29-39.

Poitou, J.P. (1995). Documentation is Knowledge: An Anthropological Approach to Corporate Knowledge Management. Proc. of

ISMICK'95, Compiègne, France, pp. 91-103.

Poitou, J.P. (1997). La Gestion Collective des Connaissances et la Mémoire individuelle, in Connaissances et Savoir-faire en entre-prise, ed. Hermès, p 157-178.

Pomian (1996). Pomian, F. Mémoire d'entreprise, techniques et outils de la gestion du savoir. Ed Sapientia.

Schreiber, G., Wielinga, B., Akkermans, H. , van de Velde, W. and Anjewierden A. (1994). CML: The CommonKADS Conceptual

Modelling Language. In L. Steels & al, eds, A Future for Knowledge Acquisition: Proc. of EKAW'94, pp. 1–25, Hoegaarden,Belgium, Sept. 1994. Springer-Verlag, LNAI n. 867.

Simon, G. (1996). Knowledge Acquisition and Modeling for Corporate Memory: Lessons learnt from Experience. In B. Gaines,

M. Musen eds, Proc. of KAW'96, Banff, Canada, November, pp. 41-1 41-18.

Simon, G. (1997). Modèles et méthodes pour la conception des mémoires d'entreprise. Le système DOLMEN : une application en

métallurgie. PhD Thesis, Université Henri Poincaré, Nancy I, Janvier.

Simon, G. and Grandbastien, M. (1995). Corporate knowledge: a case study in the detection of metallurgical flaws. Proc. of

ISMICK'95, Compiègne, France, pp. 43-52.

Simone, C. (1996). Collective Memory from Everyday Work. In Y. Wærn (1996).

Sorensen H., O'Riordan, A. and O'Riordan, C. (1997). Profiling with the INFOrmer Text Filtering Agent, Journal of Universal

Computer Science, 3(8), 988-1006.

Steels, L. (1993) Corporate knowledge management. Proc. of ISMICK'93, Compiègne, France, pp. 9-30.

Tennison J., and Shadbolt, N. R. (1998). APECKS: a Tool to Support Living Ontologies, Accepted at the 11th Workshop on

Knowledge Acquisition, Modeling and Management (KAW'98), April, 1998.

Thomas, P. (1996). Introduction: CSCW Requirements and Evaluation. In P. Thomas (ed), CSCW Requirements and Evaluation,

London: Springer, 1-9.

Tourtier, P.-A. (1995). Analyse préliminaire des métiers et de leurs interactions. Rapport intermédiaire du projet GENIE, INRIA-Dassault-Aviation.

Trigano, P. (1994). Automatic Indexation and Knowledge Storing. Proc. of ISMICK'94, Compiègne, France, pp. 223-235.

Urban, G.L. and E. von Hippel (1988). Lead user analyses for the development of new industrial products. Management Science,

1988. 34(5): p. 569-582. Also in http://iir1.uwaterloo.ca/MOTW96/readings96/urban_vonhippel.html

Uschold, M., King, M., Moralee, S. and Zorgios, Y. (1996). The Enterprise Ontology. March. http://www.aiai.ed.ac.uk/~entprise/

enterprise/ontology.html

Vandenberghe, L., and de Azevedo, H. (1995). Multi-Agent Systems & Knowlege Capitalisation: an Overview. Proc. of

ISMICK'95, Compiègne, France, pp. 61-68.

Van der Spek R. (1994). Towards a methodology for knowledge management. In J. P. Barthès ed., Proc. of ISMICK'94, Compiè-gne, October 1994, pp. 93-102.

Van Engers, T. V., Mathies H., Leget J., and Dekker, C.C.C. (1995). Knowledge Management in the Dutch Tax and Customs

Administration: Professionalisation within a Knowledge Intensive Organization. Proc. of ISMICK'95, Compiègne, pp. 71-80.Van Heijst, G, Van der Spek, R., and Kruizinga, E. (1996). Organizing Corporate Memories. In B. Gaines, M. Musen eds, Proc. of

KAW'96, Banff, Canada, November, pp. 42-1 42-17. Also in http://ksi.cpsc.ucalgary.ca/KAW/KAW96/KAW96Proc.htmlWærn, Y. (1996). Collective Learning and Collective Memory for Coping with Dynamic Complexity - Co-tech Workshop at

ECSCW 95. SIGCHI Bulletin, 28(3),Also in http://www1.acm.org:82/sigs/sigchi/bulletin/1996.3/waern.htmlWisner, A. (1997). Anthropotechnologie: vers un monde industriel pluri-centrique. Toulouse: Octarès.

20