Knowledge Acquisition and modeling for corporate memory Knowledge Acquisition and modeling for corporate memory: lessonslearnt from experienceGaële Simon CRIN/CNRS Université Henri Poincaré - Nancy 1 B.P 239 54506 Vandoeuvre-lès-nancy CedexFRANCEEmail : Gaele.Simon@loria.frAbstract In this paper, we describe important steps of the knowledge capitalizationprocess we are working on, in a metallurgical domain. From this particularand practical experience, our purpose is to focus on general characteristicswhich seem to be reusable for other knowledge capitalization systems. Wewould like, particularly, to put emphasize on specific constraints linkedto the design of this kind of systems, in order to make general methods,techniques and tools emerge allowing to answer to these requirements.1. PROJECT CONTEXT 2. EXPLOITING THE EXISTING DOCUMENTS3. THE SYSTEM ARCHITECTURE 4. CONSEQUENCES ON KNOWLEDGE REPRESENTATIONAND MODELING 5. HELP AND EXPLANATIONS IN A KNOWLEDGECAPITALIZATION SYSTEM 6. SYNTHESIS ON THE KNOWLEDGE ACQUISITIONPROCESS 7. CONCLUSION REFERENCES1. PROJECT CONTEXT The capitalization experience we are working on is part of a largerproject of corporate memory (Macintosh, 1994) and knowledge capitalizationin a firm producing steel. This firm wants to save and capitalize its knowledgeand its know-how concerning the production descriptions of the producedsteels and, also, the metallurgical defects encountered during these productions.Indeed, if they have a real corporate memory at their disposal, new expertswill be able to better understand the choices which have been done by theirpredecessors. Moreover, to maintain the global quality of the production, it seemsinteresting for this firm to be able to exploit past mistakes or successesin order to reduce, as much as possible, the risk of mistakes in the designof new steels. In this paper, we mean by the expression corporate memory , " astructured set of knowledge related to the firm experience in a given domain ". We mean by the expression knowledge capitalization, " theprocess which allows to reuse, in a relevant way, the knowledge of a givendomain previously stored and modeled, in order to perform new tasks ".The new task to perform corresponds to the purpose of the capitalizationprocess. In the context of our project, the knowledge previously storedcorresponds to the corporate memory. So, the problem we are interestedin, is to propose an implementation for the corporate memory containingknowledge on steel productions and metallurgical defects, and to associateto this corporate memory a reasoning mechanism in order to facilitate itsuse. Up to the beginning of the project, no real structured information existedconcerning the steel production process descriptions or generated defects.The main reason of this lack comes from the fact that the firm is composedof several geographic sites which have their own know-how. These sitesdid not have really the opportunity to compare their knowledge that iswhy nobody can have a global view of the knowledge of the firm concerningsteel productions and defects. In order to answer to this lack, a first step of the project consistedof the creation of common synthesis documents for all the sites.The initial aim of these documents was to allow experts from the differentsites to describe their knowledge in the same format in order to be able,afterwards, to compare them more easily. The format of these documentshas been designed in order to ensure that it could allow to homogenizethe knowledge of the different sites. Two kinds of documents were created: a first kind allows to describe steel production processes and the otherkind allows to describe already known defects. The chosen format reflects,in fact, the firm point of view concerning the described concepts.The set of synthesis documents represents a first step of the corporatememory which must be implemented. It is important to notice that thesedocuments have not been specifically designed for a future computer implementationof the knowledge they contain. The second step of the project consists in exploiting and broadeningthe synthesis effort begun by the firm during the first step. This secondstep is the object of our work and our study (Simon and Grandbastien 1995).It consists, in one hand, in proposing models allowing to implement thecorporate memory in a computer system, using the set of synthesis documents.In the other hand, our study must provide in the same system one or severalcapitalization processes allowing to use the implemented corporate memory.A first capitalization purpose has been chosen and defined : it is calledthe defects detection. It consists in determining whether a givensteel production process description can generate defects already describedin the corporate memory.In this paper, we present for each important step of the design of thesystem, which general characteristics or constraints have been pointedout, which methods or tools we have used to answer to these constraintsand which parts of these methods or tools seem to be reusable for othercapitalization systems. As a consequence, first of all, we present howthe existing documents can be exploited at the beginning of the systemdesign. Then we focus on the architecture of the system and the differentkinds of users which seem to be necessary in order to make theses systemsrun. In a third part, we introduce concepts of help and explanations inthis kind of systems. We present which kind of help can be proposed andfor which users. Finally we come back to the knowledge acquisition processand show that the method we have used can be generalized. We show thatthis knowledge acquisition method can be not only used for the main capitalizationtask but also for the modeling of the different help and explanations.2. EXPLOITING THE EXISTING DOCUMENTS The aim of this section is to point out the general characteristicsof existing documents as we have perceived them in the context of our projectand which seem to be shared by most of computerized corporate memoriesbuilding experiences. We present which knowledge acquisition process hasbeen used to answer to these characteristics. We will see that this processis composed of three main steps. In the next sections, we show how thesecharacteristics have also influenced the system design and the modelingof the necessary knowledge. Knowledge typologySeveral knowledge classes can be distinguished according to theirrole or their source. Concerning their role, one can distinguish capitalizationtask specific knowledge or, at the opposite, general knowledgewhich must be integrated in the corporate memory. For example, in our project,knowledge allowing to evaluate occurrence risks of defects can be distinguishedfrom general knowledge describing those defects.Knowledge classes can be also distinguished according to their source.A large part of them, useful, both for the corporate memory and the capitalizationprocess can be found in the existing synthesis documents which are describedlater. But we show that, in spite of the synthesis effort of the experts,a set of knowledge stay in the experts' heads and must be pointed out byusing interviews.Finally, one must distinguish knowledge really useful for the corporatememory design or the capitalization process from others. Indeed, althougha corporate memory is supposed to be very general, it concerns a precisedomain of work in the firm and it is built in a particular intention. Soa coherent set of interesting knowledge must be found. This knowledge mustbe liable to help the users of the corporate memory to better know thechosen domain and to allow to implement various capitalization tasks.Why and how to exploit existing documents?In the context of the building of a corporate memory, the acquisitionprocess must begin by the exploitation of existing documents for threemain reasons. First of all, even if these documents are not specificallydesigned for the purpose of the system to be built, it is impossible toask experts to express again knowledge they have already described and,in part, modeled. Secondly, it is crucial that the way the knowledge ofthe corporate memory is implemented is not too far from the way it is expressedin the synthesis documents in order to avoid the experts to be bewildered.Last, but not least, these documents are a very good mean to facilitatethe communication with experts; so they must be studied by the knowledgeengineer very soon in the acquisition process.Using documents as a modeling frameThe first use of the documents can consist in exploiting their structure.Indeed, the documents, produced by a firm in order to save its knowledge,are often more or less implicitly structured. If this structure is notso obvious, the first task of the knowledge engineer must be to make itemerge. It is important to take this structure into account because itoften represents a first step of modeling of the concepts to beincluded in the corporate memory. In the context of our project, the structure of the documents describingdefects includes information blocks bound by semantic links, the type ofwhich is often causal or temporal. Experts were in charge of filling upthe different information blocks according to the steel production processor the defect to describe. Each information block is identified by a fixedtitle which is supposed to reflect the semantic of the block. Figure 1shows a part of a document describing a defect containing some of the blocks.The semantic of these blocks is the following : ¨ Problem : this block describes in a few words which kindof defect is treated in the document. ¨ Aim : this block describes what are the consequences ofthe defect and particularly the percentage of occurrences of defects ofthis kind which may be allowed. ¨ Analysis : this block is composed of three other blocksand describes the mechanisms of occurrence of the defect. The " historic "bloc contains a short history of the actions which have already been achievedin order to understand or correct this defect. The block called " financialimportance " presents the financial consequences of the defectoccurrence and profits which can be hoped if the defect can be eliminated. Using documents as acquisition gridsIn a second time, documents need to be exploited in more details inorder to focus on their precise semantic. This new use of the documentsshows generally that they contain a lot of implicits which must be clearlyspecified. That is why, after having used the documents' structure as amodeling frame, the knowledge engineer needs to base his work on a setof interviews with experts who have filled up the documents. During eachinterview, the synthesis document can be used as a sort of acquisitiongrid which would be already filled and from which the knowledge engineercan discuss with the expert.The intensive use of the existing documents during the acquisition processhas several advantages and disadvantages. A first advantage is to allow to obtain rapidly a first model of theknowledge to represent in the system for the capitalization task. As aconsequence, when interviews begin, the knowledge engineer has alreadya model to work with and to show to the experts. A second advantage is that the documents can be used as a support duringinterviews. As a consequence experts know what to speak about with theknowledge engineer. So the communication is facilitated.The intensive use of the documents implies some limitations too. Firstof all, they often contain a lot of implicits. Indeed, to summarize a partof knowledge in a document implies to make choices. Secondly, even if acommon format is defined for all the documents, they are not homogeneous.Indeed, for example in our project, each expert has its own understandingof the semantic of the different information blocks of the format and,as a consequence, we don't always find the same kind of information inthe same block. The next part presents which kind of architecture can be used to representknowledge capitalization systems and which kind of users are concernedby those systems. 3. THE SYSTEM ARCHITECTURE In this part, we try to show that the kind of systems we describe inthis paper implies an architecture identifying several kinds of specificusers. Figure 2 shows the general architecture we think to be suitable to manycapitalization systems. First of all, this schema shows that, in this kindof systems, three general types of users can be distinguished. Each typeof user is provided with a dedicated interface allowing him to manage differentmodels of knowledge. Different kinds of usersThe first type of user is called the " final user ".This user is supposed to be interested in the capitalization side of thesystem. The main purpose of this user is to submit a new problem to thesystem which will be solved by a capitalization process. That is why twointerfaces called " problem description interface "and " results consulting interface " are associatedto this user. The first one helps the user to describe the data of the problem hewants to be solved. This description is used by the system to build the" problem model ". The design of this interfacedepends on the capitalization aim of the system. In the context of ourproject, the aim of the capitalization is to detect defects from a steelproduction process description. So the associated interface allows theuser to describe and to give values to the set of metallurgical parameterslinked to the type of steel to be produced.The second interface associated to this user allows him to see the resultsof the capitalization process used by the system in order to solvethe problem. In the context of our project, the result is a list of defectscoming from the corporate memory associated to a risk of occurrence. Usingthe proposed interface, the user can see, for each detected defect, thedetails of the evaluation of its risk of occurrence.The second type of user is called the " administrator ".This user is in charge of the management of data bases used by the systemto achieve the capitalization aim. These bases are supposed to containbasic knowledge of the capitalization and corporate memory domain. As forthe first type of user, a specific interface is associated to this user,allowing him to consult and to modify the data bases. In the context ofour project, those data bases contain, particularly, the description ofall metallurgical parameters used by experts or engineers in their work.The third type of user, called " domain expert "plays a major role in the system. Indeed, he has to enrich and maintainthe " corporate memory " part of the system.The main aim of the acquisition step, described in the second part of thispaper, was to find a model allowing to model this corporate memory. Asa consequence, a fourth interface must be defined in the system, allowingthe domain expert to manage the corporate memory using this model. In thecontext of our project, the corporate memory consists essentially of defectdescription using the defect model designed during the acquisition process.In that context, we do not have one domain expert but a set of expertswho have filled up the synthesis documents describing metallurgical defects.The associated interface allows each expert to create or modify new orexisting defects. To describe his defect, the expert uses a little modelinglanguage based on the defect model showed later. He can also add commentsto his defect description in order to explain his modeling choices.The capitalization taskThe model called " capitalization model "is the only one which is not accessible to users. This model contains knowledgeand methods allowing to exploit the corporate memory, using databases and the problem description, in order to generate results expectedfrom the capitalization process. Each time the type of capitalization taskchanges, only this part of the system will have to be modified. In the context of our project, the capitalization model uses case-basedreasoning (Kolodner, 1992) (Aamodt and Plaza, 1994). Indeed, as it is presentedin the next section, the corporate memory is represented by a case base,each case representing a defect of a synthesis document in terms of conceptsof the defect model. Case-based reasoning is a particular kind of reasoningby analogy. It consists, in general, in solving a new problem by usingcases already solved and collected together in a base. The first step consistsof a selection of a subset of cases from the base which are judged to berelevant to the problem to solve. In a second step, each case of the subsetis compared to the problem in order to calculate a similarity measurebetween them. Generally, a last step consists in adapting the solutionof the case the most similar to the problem in order to obtain a solutionfor this problem. In the context of our project, only the two first stepsare used. The defects are considered as " cases already solved "and the steel production process to analyze is considered as the problemto solve. As a consequence, in our system, the capitalization process consistsin searching in the case base, the set of defects the most similar to thesteel production process proposed by the final user (for further details,see (Simon, 1996)).Why have we chosen this method for the capitalization process? The mainreason is that, during the knowledge acquisition step, no general methodconcerning the detection know-how could be obtained or modeled. Indeedexperts do not really perform this detection task when they design a newsteel production process because it is too complex and too long. It isprecisely because experts do not perform completely this task each time,that the system is useful. It is a way to extend experts' memories in orderto reduce their mistakes. Case-based reasoning techniques are very wellsuited to situations where no other general method exists to perform aparticular task and where a collection of cases already solved exists.That is why we have chosen to use it, considering a defect as a descriptionof a bad steel production process in which only parameters involved inthe defect occurrence are given. Finally, the last module of the architecture called " capitalizationresult " contains a modeling of the results of the capitalizationprocess. The aim of the interface called " results consultinginterface " is to present this modeling in a suitable way tothe final user. In the context of our project, this module provides theusers with a list of the defects which are considered by the system ashaving a risk of occurrence if the proposed steel production process isreally used. A quantitative value of this risk, calculated by the system,is associated to each detected defect.In the next section, we show which consequences, in terms of modelingand knowledge representation, this kind of system and the associated architecturemay imply. 4. CONSEQUENCES ON KNOWLEDGE REPRESENTATION ANDMODELING In this section, we show which type of constraints must be taken intoaccount when designing a knowledge capitalization system based on a corporatememory and which consequences these constraints imply for knowledge modeling.We show, too, how they have been taken into account in our specific context.Architecture of the corporate memory : using casesFirst of all, we would like to emphasize the choice which has been madeconcerning the corporate memory representation. This corporate memory,that is, as far as our project is concerned, the set of metallurgical defects,has been represented using a collection of cases. We think thatthis kind of representation for the corporate memory is general and canbe reused for other knowledge capitalization systems because it answersto a set of general constraints induced by all capitalization systems (Caulierand Houriez, 1995).First of all, a firm has very often as its disposal a collectionof past experiences for which the solutions are known and which canbe easily transformed into cases. Moreover, a corporate memory always evolveand can consequently never be built in one time. A representation interms of cases allows an incremental design of the corporate memoryby adding progressively new cases. Finally, a third characteristic of thiskind of system is very often the scattering of the expertise necessaryto the building of the corporate memory to be implemented. By that veryfact, a corporate memory is the result of the global experience of thefirm in its domain. And this experience is the result of the union of theknowledge of all the experts and engineers of the firm. A representationof the corporate memory using a collection of cases allows to answer tothese characteristics because it ensures the locality of the modificationsof this memory.In the context of our project, the corporate memory is partly composedof defect descriptions. It is, consequently, represented by a collectionof cases, each of them representing a defect. Figure 3 presents the casemodel allowing to represent a given defect. The case base can be used indifferent ways according to the purpose of the chosen capitalization task.As far as our project is concerned, the purpose of the capitalization isthe defects detection and the chosen technique to achieve this task isthe case-based reasoning. One could also imagine to use other reasoningtechniques such as machine learning relying on neural networks for otherkinds of capitalization tasks.We have not evoked the problem of the organization of the casebase. In our system, the case base is not yet organized, all cases areat the same level. In a general context, the choice of the case base organization,for example a hierarchical one, can be a way to model a part of the expertise.Different levels of knowledge representation and modelingThe same knowledge environment is supposed to implement a corporatememory and propose capitalization mechanisms, this implies the need tomodel some concepts at different levels. In the context of our system,the defect concept is represented at three different levels. Within thecase model, there exists two levels : the level called " generaldescription " and the level called " causal description ".The first one describes the defect at a very general level with its name,the general conditions under which it has occurred, its physical appearance...This first level contains knowledge which can be termed as " surfaceknowledge ". Knowledge is often expressed as texts in this level. The second level presents the defects from the formation point of viewand points out the different metallurgical involved parameters and theirmutual interaction. This level contains more precise knowledge than theprevious one. In this level, knowledge is structured into objects and modeledthrough a specific modeling language defined by the knowledge engineer.This language allows to formalize the knowledge. Finally, there existsa third level of representation of defects in the capitalization model.Indeed, as seen previously, the capitalization model uses the corporatememory and consequently its cases in order to generate expected results.So this model must contain further knowledge, concerning defect descriptions,allowing to make the descriptions contained in the cases more dynamic andoperational. In the context of our project, the capitalization model contains,for example, a representation of the " parameters combinations "(figure 3) in terms of mathematical functions. For any kind of capitalizationtask, these different levels of knowledge representation can be found inmost capitalization systems: a " surface " level, amore precise and generally causal level and a dynamic level bound to thekind of capitalization task.Evolving knowledgeA major characteristic of the knowledge capitalization systems is thefact that the knowledge they contain is always evolving. The architectureand the knowledge representations must be chosen to allow this evolution.This implies that such a system must be the more generic as possible atevery levels :¨ The basic knowledge of the system : the basic knowledgeof the system, in our project the metallurgical parameters, is knowledgeused in the corporate memory description and in the problem description.We have shown that, in the proposed architecture, this knowledge is storedin separated data bases. This allows to obtain a first level of generalityof the system. Indeed, the enrichment of these bases will immediately and,without any other modification, enlarge the system " know-how ".For example, in our system, the addition of new parameter descriptionsin the bases implies that they are automatically put at the users' disposalin order to describe steel production processes or defects. Changing thosebases could be a way to specialize the system for a particular geographicsite or specific domain in the steel design. ¨ The capitalization model : Again, the capitalization modelis a separate model in the general architecture and only uses the otherone according to its own knowledge. One can hope, consequently, that whenthe kind of capitalization task will change, this model will be the onlyone to adapt. Everything else in the architecture is designed in orderto remain the same from a capitalization task to an other. ¨ The corporate memory model : the corporate memory is implementedwith a case base. This representation, as seen previously, answers, inpart, to the problem of the evolution of knowledge. But it is not enough.The case model must also be as general as possible in order to be usedas a corporate memory support and to allow different kinds of capitalizationtasks. For that, it is crucial that the defect model must be declarative.Finally we would like to point out the fact that the case-based reasoningtechnique specifically used, in our context, as the capitalization modelanswers very well to the problem of the knowledge evolution. Indeed, thedefects base will always evolve and each case description will also evolveas their understanding by the experts will become better and better. Asthe detection task is based on the evaluation of a similarity measure betweenthe analyzed steel production process and the defects, new defects willautomatically be taken into account in the evaluation. So the detectionmechanism will evolve itself. 5. HELP AND EXPLANATIONS IN A KNOWLEDGE CAPITALIZATIONSYSTEM Until now, we have only been interested in knowledge concerning thecorporate memory or the capitalization task. In this section, we deal withknowledge necessary to allow to propose some help to the users of a capitalizationsystem. Indeed, it has been shown that integrating explanation functionswithin a given system needs kinds of knowledge which are different fromthose necessary for the problem solving task (Clancey, 1983). Differentkinds of help and explanations functions, which can be considered in aknowledge capitalization system, are presented. The last step of the firstknowledge acquisition process for the main capitalization task, consistingof a sequence of tests with a prototype with an expert, allowed us to identifyexplanation and help needs at different levels. The needs are differentaccording to the different kinds of users (see section 3) which are considered.More generally, we think that four kinds of help or explanations can beidentified in such a system : ¨ to help the different users to use the software ¨ to help the final user to interpret the results producedby the system in terms of corresponding synthesis documents. ¨ to suggest to the final user some modifications of hisproposed problem. In our context, it can consist in suggesting modificationsin the proposed steel production process in order to eliminate or reducethe effect of the detected defects. ¨ to help the domain experts to use the case model in orderto be able to enrich and maintain the corporate memory.The first three kinds correspond to the traditional help and explanationswhich can be found in expert systems or CBILEs (Computer Based Interactive Learning Environments).The first type corresponds to the on-line help which can be found in everysoftware. The second one corresponds to the help for a better understandingof the results produced by expert systems. The third onf help which can not be easilycharacterized in comparison with already known kinds of help. The mainpurpose of this help is to allow the domain expert, in our context thedefect expert, to be as familiarized with the case model as possible inorder to facilitate the modeling of his own cases. This kind of user isnot supposed to use the system very often and, as a consequence, is notspecially very informed about how the system runs. He is not the finaluser. This characteristic must be taken into account in the design of thehelp for this user. This help can be given in three ways : ¨ a synthetic presentation of the model produced by the user inorder to make him understand what are the consequences of his choices. ¨ an explanation of the different components of the defect modeland their role (relying on KADS models for example). ¨ a dynamic help to the modeling task of a new case of thecorporate memory, in our context a defect.In the following presentation, we are only interested in the explanationsof the results produced by the system for the final user and in the dynamichelp to model for domain experts. For each of them, we present their purpose,their content and the knowledge acquisition method which can be used inorder to generate and propose those help and explanations.Explaining the results of the capitalization process to the finaluserThe final user of a capitalization system is not an expert of the corporatememory domain but has a minimal set of knowledge. In our context, he canbe a metallurgist engineer. As a consequence, the purpose is neither toexplain the basic concepts of the domain nor to explain how the systemhas produced the results because those results contain the set of informationallowing to understand the similarity measure calculus. But this user seesthe corporate memory only through the cases' models produced by domainexperts. He can not see choices on initial knowledge those experts hadto do to achieve their models. The idea of the proposed help is to allow the final user to interpretthe results of the system. In fact, what we want to propose to the finaluser is a generation of explanations which could be given by a real domainexpert seeing the problem of the final user. It can be considered as similarto the explanations generated by the system REX (Wick and Thompson, 1992).A way to do that is to automatically reproduce comments written by expertsin their synthesis documents which, often, correspond to the type of explanationwe are interested in. In the context of our project, these explanationsshould help the final user to interpret the risk of occurrence of a detecteddefect in terms of metallurgical, chemical or mechanical mechanisms comingfrom defect descriptions. This interpretation should help him to reconstructthe initial description of the defect before the domain expert modeling.The aim is somehow to propose to decompose the knowledge the domain experthas modeled.Helping domain experts to use the case model in order to enrich thecorporate memoryThis kind of help is relatively new in comparison with the existingkinds of help because it is linked to the enrichment of the knowledge ofthe system. Indeed, the corporate memory will always evolve. As a consequenceseveral persons, the domain experts, will have to maintain and modify thismemory. Those users do not use the system in an intensive way. So theydo not know completely how the system runs and which knowledge it uses.They only know a part of this knowledge : the cases they have to model.In the context of our project, more than ten domain experts will have tomaintain the corporate memory. The role of this kind of user is crucialbecause the quality of their model have a direct influence on the qualityof the results of the system. So, the main characteristics of thiskind of user, to take into account for the generation of explanations,are the following : ¨ He is an occasional user. ¨ He does not use the system to solve any problem. ¨ He is a domain expert who is in charge to put a part of his knowledgeat the system's disposal via the corporate memory. ¨ He does not know exactly how the system runs.As the role of this user is so crucial, it seems important to help himto achieve his task of knowledge enrichment and modeling. This help canbe achieved by dynamically guiding the expert during the use of the modelaccording to his needs of modeling. The modeling task can be seen as specificproblem solving activity. So this kind of help could seen as a more traditionalhelp to problem solving. The specific difficulty here, is that the systemdoes not know what the user wants to model. So the system must begin, whenhelping the expert, by identifying the type of difficulty encountered bythe expert, for example by questions. In fact, the content of this kindof help can be seen, in part, as the set of questions asked to the expertsduring the acquisition process in order to design the model. The purposeof the help is to give a solution, using case model components, to themodeling problem expressed by the expert.Integration of these helps and explanations in the capitalizationsystemGeneration of kinds of help and explanations described previously willneed to exploit the corporate memory again. The main difference betweenthis exploitation and the one made for the detection task is that the generationof explanations needs also knowledge coming from the results produced bythe detection task. So the generation of each kind of help or explanationwill be implemented in new modules, data entry of which are in one handknowledge coming from the corporate memory (as for the detection task)and in the other hand knowledge coming from the results produced by thedetection task . 6. SYNTHESIS ON THE KNOWLEDGE ACQUISITION PROCESSThe knowledge acquisition step of a system design is no longer seenas a simple transfer of the expertise from experts to abstract structures.It is now seen as a real problem solving task the purpose of which is todesign models reflecting the collected knowledge (Aussenac-Gilles, Krvineand Sallentin, 1992).In this section, we present a general knowledge acquisition method adaptedto the context of capitalization systems design within a firm. As in otherknown knowledge acquisition methods like KADS (Wielinga, Schreiber andBreuker, 1992; Schreiber, Wielinga and Breuker, 1993) or CERISE (Vicat,Busac and Ganascia, 1993), the acquisition process is perceived as a cyclicprocess of progressive knowledge elicitation and modeling. In the methodwe propose, this process is composed of three main steps, the last twoones being cyclic. What we call "acquisition step" is a partof the global acquisition process allowing to obtain an intermediate stateof the knowledge model. This method must be used for each capitalizationtask to be included in the system and allows to obtain and model the knowledgeneeded for this task. What we propose is a general approach to drive theacquisition process for a capitalization system. But for each step of themethod, we don't specify any modeling language. Existing ones, like theKADS expertise model, can be used.Figure 4 presents a schema which summarizes the proposed method. Inthis schema, each step, symbolized by a circle, is associated to four otherselements. First of all, each acquisition step has an input and an outputwhich correspond to different states of the final models to be obtained.The input of the step i is the output of the step i-1.  As a consequence this method proposes an incremental developmentof the models. Each step is also associated to one or several supportsto be used during the step. It can be existing documents, a prototype orthe experts. For some steps, as the first one, supports are fixed, forothers they must be specified according to the capitalization task forwhich the method is used. Finally, each step is associated to a generalpurpose which can be stated precisely for a specific acquisition process.The first step consists of the exploitation of the structure of theexisting documents. We have already dealt with this exploitation in thesecond part of this paper. The aim of this first step is to categorizethe knowledge that is to say which kinds of knowledge must be taken intoaccount in the system. The support of this step is essentially the setof documents. This step produces a result called " knowledgestructuration " that is to say a first organization of the knowledgebut without any choice concerning the way they must be modeled. The second step consists of interviews with domain experts. The generalpurpose of this step is to deepen the knowledge included in the "knowledgestructuration" obtained in the first step in order to be able to modelit precisely. This "knowledge structuration" is used as a modelingframe during this second acquisition step. The support of this step is, at least, the set of experts concernedby the corporate memory domain. But other supports can also be used accordingto the purpose of the chosen capitalization task. The result produced bythis step is a first model of the knowledge to be included in the corporatememory in order to achieve the capitalization task. This step consistsof a sequence of several interviews, each of them producing a new stateof the model taking into account the results of the discussion betweenthe knowledge engineer and the last expert. This new state is used as thebeginning point for the next interview of the sequence. It correspondsto the fact that, in the context of the design of a corporate memory, theknowledge is always disseminated among several experts. Each expert workon a different case (defect) which minimizes the possible contradictionsbetween them. As a consequence, the method propose an incremental andprogressive refinement of the model. In the context of our project, we have chosen a set of five synthesisdocuments describing defects. These defects were judged by experts as beingrepresentative of the set of existing defects descriptions diversity. Thepurpose of each interview was to try to model each of these defects, takinginto account the "knowledge structuration". As we have shownit in part two, synthesis documents were also used as support of this step.The last step is called " prototyping and tests ".It consists in testing the model, obtained at the end of the previous step,with the help of a prototype implementing this model. The purpose of thisstep is to verify if the previous model is complete, that is to say allowsto represent any element to be included in the corporate memory in orderto perform the capitalization task. Indeed, at the end of the second step,a model is produced allowing to represent a subset of cases. The knowledgeengineer has then to ensure that this model is general enough. So this purpose can be called, as in the schema, " the knowledgeenrichment ". To achieve this enrichment, the support of thestep must be at least the prototype and the experts but additional supportscan be used. This last step consists of a sequence of trials with different experts,each trial enriching progressively the model. The final result of thisstep and of the acquisition process is a refined model. Between each test,the evolution of the model must be taken into account in the prototypein order to allow next experts to have a view of the work already donewith the other ones. It is important that the experts working in the laststep are not the same as those of the second step because the third acquisitionstep must give information not only about the generality of the model butalso about its easiness of use. As a consequence, it is impossible to workwith experts who were involved in the building of the model. In the context of our project, the defect model has been implementedwithin a prototype which performs the chosen capitalization task, thatis the defect detection, but only on a little number of defects. Each sessionwas driven in the same manner : the aim was to use the prototype with theexpert in order to represent a new defect in the model. Each time the workwas based on the corresponding synthesis document. As a consequence, synthesisdocuments have been used as a further support of this acquisition step.Twenty sessions with thirteen different experts were organized. At theend of this third acquisition step, we have obtained a final model of thedefect concept, available both for the corporate memory and for the capitalizationtask. We think that this general method for knowledge acquisition can be reusedfor any capitalization task. Each time the supports and purposes of eachstep will have to be specified according to the type of the capitalizationtask. A first advantage of this method comes from the fact that the interviewsbegin only in a second time. It helped us to center those interviews. Indeed,first of all, each interview was not centered on the general know-how ofthe expert but on a concrete case he had already described in a document.So the set of knowledge to be covered during the session was small. Finally, the aim of each session was precisely defined : to be ableto use the model in order to represent a given defect. It helped to bealways centered on the set of knowledge we were interested in.As a second main advantage of our method, we would like to underlinethe major role played by the prototype during the third acquisition step.Indeed, this experience shows that it is only when the future users ofa system see this system running that they really describe what they expectfrom the system, give advice and have a precise idea of what kind of knowledgewe want them to transmit. During the last step of the acquisition process,we have collected new information, particularly about defects, which couldnot be obtained during the second one. It can be explained by the factthat, in the second step, they were guided by our questions during theinterview. In the last step, the task to be achieved during the interviewswas the same: to model a given defect. But this time, the experts did nothave to answer to our questions but to use the prototype. The prototypeallowed them to understand really how the model is used by the system toachieve the capitalization task. As a consequence, they understood thatthere was a need to add new kind of knowledge in the model in order toimprove the system's results.Moreover, we would like to underline an other difficulty of the acquisitionprocess in the general context of capitalization systems. In this kindof systems, knowledge to be modeled and capitalized is often evolving.As far as knowledge acquisition step is concerned, it implies several back-trackduring the process. In the context of our project, a defect descriptionis always evolving. Each time a new defect is discovered, the notion ofdefect may have been modified. As a consequence, it was very importantto choose a " good " first set of defects for the modeldesign in order to minimize the number of back-tracks.In a capitalization system, different kinds of knowledge must be represented: for the corporate memory, capitalization tasks, explanations... So severalknowledge acquisition processes must be performed. As a consequence, itis important to try to unify these acquisition processes as much as possibleby using a common approach. The method we have presented tries to answerto these requirements. In the context of our project, it has been usedtwice. It was used a first time for the acquisition of knowledge concerningthe detection task. And it is now used a second time for the acquisitionof knowledge concerning the explanations generation. Each acquisition processcorresponds to the process described by the method, and is consequentlycomposed of three main steps. We now show how the method has been usedfor the explanation part.Knowledge acquisition processes for help and explanationsThe first step of the method allows to obtain the kinds of knowledgenecessary to generate those explanations. The support is the set of synthesisdocuments and particularly the comments written by the experts in thosedocuments (see before).The second step corresponds, in fact, to the third step of the acquisitionprocess for the main capitalization task, in our context the defect detection.As a consequence, the support is a prototype without any explanation functionand the experts. This step allows to specify precisely the need of expertsfor explanations. By observing them using the prototype, one can collectinformation on their difficulties in order to model them. So this stepallows to produce a first model of these difficulties. The third step, as for the main capitalization task, allows to enrichthe previous models in order to obtain complete models. Here, the supportis again a prototype but, this time, with explanation functions correspondingto the models obtained at the end of the second step. This time the testswith the prototype must be specifically dedicated to the problem of explanationswhich was not the case with the previous tests of the previous step. Indeed,an expert can not express his needs of explanation in general. So, it isnecessary to begin by observing him using a prototype and then, only ain second time, to propose explanation functions and to ask him to reactin specific sessions.The schema of figure 5 summarizes the existing relation between thethree acquisition processes. It shows they are not performed at the sametime but they are not independent. The two knowledge acquisition processesfor help and explanations are performed in the same time but, one steplater than the acquisition process for the main capitalization task. Thisschema points out the fact that the second step of the acquisition processesfor help and explanations is exactly the same as the third one for thefirst acquisition process concerning defects detection, which is the maincapitalization task. We think, indeed, that knowledge acquisition processesconcerning explanations can not be managed in the same time with the mainacquisition process (Delozanne and Carrière, 1992) because it isonly when the experts use a prototype that then can express their difficultiesand, as a consequence, their needs for help and for explanations. So, althoughactual researches try to design expert systems which have explanation functionsfrom the beginning (Bouri, Dieng, Kassel and Safar, 1990), we think thatit is very difficult to specify them at the beginning of the design. 7. CONCLUSION In this paper, we have shown that knowledge capitalization systems designimply a set of specific constraints, which are not found in all traditionalexpert systems, and which influence the knowledge modeling. Such characteristicsinclude at least an extensive use of existing documents, a continuouslyevolving knowledge base, a disseminated expertise and many expertise providers,different kinds of users.Our conclusion is that a corporate memory is altogether less and morethan a traditional knowledge based system (KBS). It is less than a KBSbecause it is not dedicated to a specific task and consequently it doesnot include know-how related to such a task. It is more than a KBS becauseit should be used for several capitalization tasks and by several kindsof users.We have adapted some general methods, techniques and tools in orderto take these constraints into account. A knowledge acquisition methodprovided for a given capitalization task has been defined. A general architecture,with different kinds of users has been also defined. Finally we have dealtwith the question of help and explanations which can be integrated in sucha system. Different kinds of help have been defined and we have shown thatfor two of them, the knowledge acquisition method can be reused in orderto obtain the new set of knowledge necessary to generate them.Present and future workOur prototype is written in C++ and Prolog and is running on PC's. Itis currently experimented for the last refinement cycles in explanationknowledge acquisition. As far as corporate memory is concerned, it remainsto observe how the case base is correctly updated and enriched by the differentexperts when they are in charge of this task.The next step will consist in trying to integrate this system in a largerinformation system in the firm. Such an integration needs to define a methodologydescribing how to use the system. More precisely, protocols have to bespecified in order to define how the corporate memory must be enriched,who must be in charge to enrich it and when to enrich it. REFERENCESAamodt, A. and Plaza, E. (1994). Case-Based Reasoning : Foundational Issues, Methodological Variations, and System Approaches, in AI Communications, Vol. 7, No. 1, p. 39-56. Aussenac-Gilles, N. and Krivine, J.-P. and Sallentin, J. (1992). Editorial : L'acquisition de connaissances pour les systèmes à base de connaissances, Revue de l'Intelligence Artificielle, Vol 6, No. 1-2, p. 7-18. Bouri, M. and Dieng, R. and Kassel, G. and Safar, B. (1990). Vers des systèmes experts plus explicatifs, in Bernadette Bouchon-Meunier (Eds.), Troisièmes jounées nationales du PRC-GDR Intelligence artificielle, Centre National de la Recherche Scientifique, HERMES publishing, p. 340-355. Caulier, P. and Houriez, B. (1995). Apports de la modélisation des connaissances et du raisonnement à partir de cas à la capitalisation et la réutilisation de connaissances, in Actes des Journées Acquisition Validation Apprentissage, Grenoble (France), p. 331-345. Clancey, W.J. (1983). The epistemology of a rule-based expert system : a framework for explanation, in Artificial Intelligence, Vol. 20, No. 3, p. 215-251. Delozanne, E. and Carrière, E.. (1992). Définir un processus explicatif, une étude de cas : la conception d'ELISE, in proceedings of Deuxièmes journées Explication du PRC-GDR-IA, Centre National de la Recherche Scientifique, Sophia-Antipolis (France): INRIA publishing, p. 195-208. Kolodner, J.L. (1992). An introduction to case-based reasoning , in Artificial Intelligence, Vol. 6, p. 3-34. Macintosh, A. (1994). Corporate knowledge management - state-of-the-art review, in proceedings of ISMICK International Symposium on Management of Industrial and Corporate Knowledge, Compiègne (France), p. 131-145. Schreiber, G. and Wielinga, B. and Breuker, J. (1993). KADS : a principled approach to knowledge-based system development, in Knowledge-based systems, Vol. 11, Academic Press. Simon, G. and Grandbastien, M. (1995). Corporate knowledge : a case study in the detection of metallurgical flaws, in proceedings of ISMICK International Symposium on Management of Industrial and Corporate Knowledge, Compiègne (France), p. 42-51. Simon, G. and Grandbastien, M. (1996). Case-based reasoning for knowledge capitalisation, to be published in proceedings of Expert Systems'96, Cambridge (UK). Vicat, C. and Busac, A. and Ganascia, J.-G. (1993). CERISE : A cyclic approach for knowledge acquisition, in Lecture Notes in Artificial Intelligence (Eds), Toulouse and Caylus (France), Proc. of the 7th European Knowledge Acquisition Workshop, Springer-Verlag, p. 237-255. Wick, M.R. and Thompson, W.B. (1992). Reconstructive Expert System, in Artificial Intelligence, Vol. 54, No. 1-2, p. 33-70. Wielinga, B. and Schreiber, G. and Breuker, J. (1992). KADS : a modeling approach to knowledge engineering , in Knowledge Acquisition, Vol. 4, p. 5-53. |
|
