Designing an Electronic Knowledge Book: How Advanced Knowledge Management Method Can Help Information Modeling.
Because of the amount of information produced and consumed by individuals and organizations, the development of information systems has exploded during the last two decades. Researchers and developers have built different types of information retrieval (IR) systems. In scientific and technical fields, organizations have their own well-defined information systems, and users of these systems often perform the same information-dependent tasks. These tasks are modeled in knowledge management projects, and the results of interviews with experts in the relevant fields are converted to knowledge diagrams describing an overview of their activity.
This article presents the design study for the electronic knowledge book, a user-centered hypermedia tool for scientific and technical information retrieval guided by knowledge diagrams. This tool allows users to navigate through knowledge diagrams that describe and map a particular domain. They can pick up elements of browsed diagrams and use them to build a query, and then send it to the information fund related to the domain using a specific form. The tool also takes advantages of the virtual documents approach.
After detailing the origin and purpose of the electronic knowledge book, we present the information access strategy it implies and the particular requirements of this strategy in terms of information modeling. We then propose a design study for this tool and present the prototype built for a unit of the French Atomic Energy Commission (CEA).
DESCRIPTION OF AN ELECTRONIC KNOWLEDGE BOOK
Individuals in organizations are constantly accessing an increasing amount of information because of the development of new information and communication technologies (NICT). According to Bracewell and Laferriere (1996), NICT refers collectively to
a series of technologies that usually include the computer and [that], when combined or interconnected, are characterized by their power to memorize, process or make accessible (on a screen or other support) and to transmit, in principle to anyplace, a virtually unlimited and extremely diversified quantity of data.... These technologies are found with increasing frequency in various forms: texts, diagrams, graphs, moving images, sounds, etc.
Technically, NICT are closely related to Internet/intranet information diffusion and to associated technologies such as HTML, XML, and Java.
One type of NICT is the virtual document. A virtual document is composed of multiple pieces of information, stored separately. With Document Type Definitions (DTDs) and stylesheets, these pieces may be combined to produce a document that is called virtual since it doesn't really exist on a storage device. This approach allows the reuse of the information components and the user-customization of the produced virtual document (Cruz and Lucas 1998; Ranwez and Crampes 1999).
In this article, we present an artifact called the electronic knowledge book (EKB), which takes advantage of virtual document technologies to help users access the information they need to perform professional tasks. The EKB is designed to be made available on an intranet network, since its content is specific to the activity of one organization. Thanks to the virtual documents approach, each instance of an electronic knowledge book is based on the same conceptual architecture. The following paragraphs introduce the origin and design process of this tool by detailing three design steps: elaborating its content, converting it into a hypermedia document, and linking it to the related pieces of information.
Defining a knowledge book
Derived from artificial intelligence and expert systems, knowledge management proposes qualitative methods for modeling human reasoning and uses those methods in various application domains, such as knowledge capitalization, formation, process re-engineering, or staff management. The Method for Knowledge Systems Management (MKSM), described by Ermine and others (1996), defines a conceptual tool for human knowledge representation and the product of that tool, called the knowledge book, which is the focus of this article.
A knowledge book is a collection of diagrams drawn according to several sets of rules, each of which represents a specific point of view. Those points of view reflect know how and knowledge related to a particular domain. The commonly used sets of rules are activities, tasks, concepts, and physical processes, but other sets of rules can be defined or used in knowledge books, depending on the project's requirements. For instance, in our pilot project, we defined an object model to distinguish the physical and non-physical artifacts that experts deal with (see "A case study in nuclear criticality safety" later in this article). Relationships exist between the different types of diagrams. For example, a concept can be related to the tasks it is used in.
The MKSM method also defines the diagram redaction process, which consists of three steps, each led by a domain external knowledge engineer.
* Interviews The engineer interviews the domain experts and asks them to explain their work and the way they do it. Each interview is recorded.
* Redaction The engineer sketches the diagrams based on the interview transcripts.
* Feedback The engineer presents the diagrams to the experts and gets their opinions. They can choose either to validate the diagrams or to ask to engineer to modify or improve them. This phase is also recorded, and the results are fed back into another redaction step.
Translating a knowledge book into a hypermedia document
A knowledge book, then, is a collection of diagrams, structured according to a definite number of semantically typed relationships. By reading the diagrams and comprehending the articulations between them, the reader can attain a global view of the domain they represent. The set of generic relationships between the diagram nodes is specified by Medini (1997). To gain all the potential benefits of the knowledge book, understanding these relationships is as important for the reader as the diagrams' content.
Originally printed on paper, these knowledge books obliged readers to continually jump from one page to another. This is why we proposed to build a knowledge book in a format that embodied all those relationships and made navigation among them easier.
A hypermedia prototype built in C+/- + by Millerat, Ermine, and Chaillot (1996) has demonstrated its suitability for the presentation of knowledge books. Because of the development of Internet and intranet networks, however, it is no longer necessary to develop stand-alone C+ + hypermedia systems. Electronic knowledge books can now be constructed using HTML and can be viewed with Internet browsers. Java technology can be used to make explicit the semantics of each relationship when the reader navigates the hypermedia (that is, thanks to pop-up windows, the reader can choose the destination type of each link starting from a node).
As with the construction of any other type of hypermedia document, transposing a knowledge book from paper to hypermedia raises some potential problems. The reader must understand the general structure of the hypermedia document and must not get disoriented by it (Conklin 1987). Indeed, as Bachimont (1996) emphasizes, if readers don't know how to find their way in the hyperlinks network, hypermedia presentation can result in the readers failing to make sense of the content. In this case, the hypermedia format is less effective than the linear one and becomes what Bachimont calls hypotext rather than hypertext. Hypermedia design, then, is a major issue for the construction of meaningful hypermedia knowledge books.
We identified two main issues in this field of hypermedia design.
When transposing linear text into hypertext, the designer splits the initial text into units that he or she thinks are pertinent and builds the hypertext structure based on this division. In the case of a knowledge book, the designer does not control these parameters: the granularity of the information is imposed-the elementary unit is of course the diagram-and the hypermedia links are defined by the semantic relations between the diagrams. The clarity of the knowledge models and the pertinence of the articulation between those models determine the intelligibility of the hypermedia document.
Numerous authors in the field of hypermedia design affirm that it is of interest to provide users with something that helps them keep track of a main thread all along their navigation of the hypermedia network. A synthesis of the existing techniques, such as conceptual maps or fisheye views, is presented by Roger, Lavandier, and Kolmayer (2000).
In the paper version of a knowledge book, readers dispose of several artifacts that help them find elements in the set of diagrams, depending on the type of diagram:
* Activities and tasks are respectively addressed by tree-structured views called functional tree and tasks tree.
* Concepts and activity performers are listed in concepts and actors dictionaries.
Because of the complexity of the relations between the domain's physical processes, a global view of these processes and relations is presented as a diagram, called the processes scenario.
These artifacts then are conceptual maps for each point of view, They can be presented at the same time as the diagrams, in multi-framed HTML pages.
Regardless of the users' interest in the knowledge models, the determining elements for the success of the hypermedia transformation of a knowledge book are designed during the knowledge modeling phase. The knowledge engineer is then responsible for the form as well as for the content of this hypermedia document during the redaction phase. In these conditions, a hypermedia designer is no longer needed and the paper-to-hypermedia transformation can be completed with a computerized tool. CEA/DTI initiated a partnership with the C-Log Company to develop a tool able to produce HTML-formatted knowledge books from knowledge diagrams. This tool is also designed to generate all the artifacts that can be automatically deduced from the diagram contents.
Recontextualizing information by linking it to the diagram collection
With the hypermedia version of the knowledge book, we produce a computerized view of the represented domain. We assume that this view can be used to describe the information system related to the domain (information system refers to all the pieces of information that can be necessary for running an organization and all the tools that allow this information to be produced, stored, and transmitted).
Our goal is then to link the hypermedia knowledge book to the domain-related information sources, so that users can locate pieces of information related to the diagram elements. Conversely, users can recontextualize within the domain activity a concept found in a document by locating the diagrams that refer to it (Chaillot and Ermine 1997).
The domain description is linked to the information system through a three-dimensional model (described later in this article) that enables the user to construct information requests. This tool is called an electronic knowledge hook (EKB). It is designed to help users build requests from diagram elements so they can find relevant documents to solve information problems.
An EKB, then, is a domain-oriented information access tool, in which users can navigate using a Web browser. This tool provides users with the ability to capture the content of a diagram or of a document, and to send it to a search engine that processes either the information system or the knowledge model's index of elements.
Pilot projects have produced prototypes that demonstrated the usefulness of such tools. Nevertheless, they support only a one-to-one, information-concept association, as shown in Figure 1. In fact, information access tasks are more complex and require more than one element to describe information needs. This is why this article proposes an information access strategy superior to the one proposed by Chaillot and Ermine (1997).
TOWARDS AN EKB-SPECIFIC INFORMATION ACCESS STRATEGY
An EKB embodies knowledge about the domain of the information system for which it was built. To take into account the users' specific needs and preferences, the EKB must reuse this knowledge during the information retrieval process. We developed a specific information access strategy that is both domain specific and user centered.
Querying vs. browsing: Different strategies for different needs
According to Pamquist and Kim (1998), individuals use an information system when they perceive "an information need, or a gap between their existing knowledge and a new situation" (p. 15). Asking a friend, a colleague, or a librarian can be considered as an interaction with an information system. These interactions continue until the information gap is filled. When users choose to use a computerized system, they are usually unsatisfied by these interactions (Conklin 1987). In the field of computerized information access systems, experts commonly distinguish two main information access strategies: browsing and querying.
* Browsing involves accessing pieces of information one by one, evaluating each piece, and navigating to another thanks to a relationship established between two pieces. Such relationships define the information system's structure. Hypermedia navigation is the most commonly used browsing strategy.
* Querying is a direct interrogation of an information system. It consists of sending a query and receiving the system response.
This gap, previously described by Belkin (1980) as an "anomalous state of knowledge" (p. 133), can he more or less precise. For Marchionini (1995), the appropriate strategy for finding needed information or filling the perceived gap depends on the precision of the information problem. For example, a well-defined information problem can be solved easily using a system that offers a querying strategy. Unfortunately, this kind of system is not as useful for solving vague problems or for answering questions that require the understanding of a global domain such as that contained in a knowledge book. In such a case, browsing the information and "berrypicking" (Bates 1989) information so as to build oneself a mental model of the represented domain is more helpful than querying.
Some authors, like Agosti and Smeaton (1996), propose to integrate these two strategies in the same system: the problem definition is completed by navigation (browsing), and the user finally uses querying to finalize the information retrieval process. This is, for example, the strategy proposed by Chaillot and Ermine (1997).
Enriching information access with user profiles
According to Chaillot and Ermine (1997), information access with the EKB is built on two dimensions: the domain knowledge models and the information system. Each user takes inspiration from the proposed view of the domain to organize his or her knowledge of this domain and of the related information system's structure. However, some projects that deal with particularly complex domains require multiple modelings to relate the points of view of experts in multiple subdoma ins.
In these projects, the EKB is intended to be used by several types of readers, each of which has a different prior knowledge of the domain, a different goal to accomplish using the EKB, and a different understanding of the same piece of information. For instance, the same object can be described in different ways by the experts who design it and by those who use it. To make matters even more confusing, the object can even have different names among different groups of users. In this case, information retrieval about the object can be focused on one of the subdomains (the object's design process or its use) or on the domain (the object) itself.
For Waern and Hagglund (1997), information interpretation processes produce a different meaning for each individual, based on their mental models of the pertinent relations within the domain. According to McNamara and others (1996), if a structured presentation of the information pieces benefits the understanding of users whose mental model corresponds with the presented structure, it can be harmful to the understanding of others.
Therefore, to facilitate information transmission from the EKB to the user, it is important to present a model of information structure (and a domain view) compatible with the user's own model. To do so, the system (for example, an EKB) must be able to select a point of view that corresponds to a particular user's mental model. To allow for this kind of customization, we chose to integrate in the EKB the notion of user profiles. These profiles define parameters, such as domain and sub-domain expertise levels or familiarity with information retrieval systems, that allow the interface to satisfy the viewing preferences and information needs of different types of users.
Combining querying and browsing: The EKB cumulative information access strategy
Browsing the knowledge diagrams allows users to gain an overview of numerous elements that can be of interest in solving their information problems. But finding a document among a large collection requires a querying information access strategy. Our goal is to bring together these two strategies. This type of EKB is designed to build query requests from elements selected in several diagrams encountered while browsing. For this purpose, the EKB provicles users with a querying form that allows them to collect one by one, as with a shopping cart, all the elements that they find relevant for solving an information problem and to submit the request when it seems to be complete.
Each selected element corresponds to a subpart of the information system. Due to the semantic variety of the existing links between the diagrams, those subparts are not necessarily included in one another, as in a classical tree-structured decomposition. Requests are built by the articulation of all the information system subparts corresponding to the selected elements.
This is where the information access strategy presented in this article differs from the one described by Chaillot and Ermine (1997). In their view, users follow an opportunistic browsing strategy in the diagram collection and the document fund, moving from one element to another. Each navigation step is entirely determined by the last chosen element. Thus, a complex information problem (that is, one related to several diagram elements) will require several requests to be solved (one for each element), each document retrieved answering to a part of this problem. According to Chaillot and Ermine, users build their own vision of the information system from the domain model: they navigate in the knowledge models to define a multicriteria request that is coherent in terms of the entire problem.
Technically, thanks to logical operators (AND, OR, NEAR.. .), users can choose their own way to associate the selected concepts and identify the information system part they want to interrogate. By default, the system links all the elements with ANDs. The logical operators between the concepts are interpreted by a rule-based system that interrogates the index corresponding to the question and to the appropriate user profiles. Of course, during the whole request construction process, users have access to the formalization of their request. Thus, they can complete or modify a request or even directly submit a query without navigating in the knowledge diagrams. Once they feel the request is complete, they submit it to the search engine (in our case, SPIRIT [TM], a multilingual, plain text natural language search engine, developed by the Technologies-Gid [TM] company), that will process the selected indexes.
Although there have been no validation tests done yet, we can guess that the success of such a hybrid information access strategy is dependent on the individual successes of both browsing and querying strategies. Indeed, this strategy will not satisfy users if they can't find relevant elements for their problems in the diagram collection or if the results of the query are deceptive in regard to the efforts they invest in elaborating the request. Even if the latter point is for information retrieval research domains rather than for EKB design (see Leloup  for an overview of existing offerings in this domain), these two points are fundamental to gain users' interest in the EKB. They lead to the problem of modeling the query definition process.
EXPRESSING AN INFORMATION NEED: THE THREE-DIMENSIONAL MODEL
The knowledge models, the user profiles, and the information system then are the three independent parameters of an information access process using the EKB (Figure 2). They will provide the EKB with the needed elements to express users' information problems and build relevant information requests. To do that, the three dimensions must correspond to predefined models, so that the system knows how to cross the given parameters. This section describes how we defined these three models before conducting a design study for this tool.
Modeling the knowledge to recontextualize information
Earlier in this article, we stated that the diagrams are written according to the MKSM knowledge management method. This method defines the sets of rules and the global architecture for a knowledge book (see Figure 3). The knowledge diagrams are used as a map of the information system and are provided as a hypermedia support to help users access information in a domain-oriented manner. The different sets of rules are
* Activities, used to describe the usual way that experts proceed to do their work
* Concepts, which describe, in a tree-structured way, all the material and conceptual objects they deal with
* Phenomena, which describe the physical processes required for understanding the domain
* Tasks, which present the strategies experts have in mind when performing activities
* Evolution of concepts, which traces the background of the currently referred concepts
* History of actions, which explains the previous evolution of activities and tasks
The success of hypermedia navigation through the collection of diagrams depends on the knowledge modeling. Each diagram must contain an elementary but meaningful piece of information. The set of possible relations among the diagrams is predefined, and maps of the hypermedia diagrams are available. During the interview and the redaction phases of the project, the knowledge engineer must keep in mind the aim of the modeling: the diagram collection must give a rich and well-balanced view of the domain, and point out the elements from which information access is useful.
Modeling the users to understand their needs
More and more systems such as operating systems or software now integrate user profiles that allow the customization of some functions. Each of these systems refers to user models (or simply features) specifically chosen with regard to the functions to be customized. For the EKB, this technique is hardly applicable since we need a pertinent user model for several purposes (diagram visualization, information access, and interface settings). It must also be reusable from one project to another. Thus, a more global model must be built, including classical features (such as users' personal navigation preferences or centers of interest in the information system) and allowing as well the definition of domain-specific features (such as the users' skill level in each modeled subdomain or particular applications).
Work on user modeling in more general disciplines such as human-computer interaction or cognitive science are more useful for our EKB design study. For Waern and Hagglund (1997), who studied user modeling in knowledge-based systems, the most frequently used characteristic is users' level of domain expertise. Waern and Hagglund also remark that for complex systems, a single expert-novice distinction is insufficient to give a fine enough description of various users' expertise levels. Allen (1997) notes that it is pertinent to pinpoint the users' knowledge level of information access systems.
In disciplines more directly related to our problem, Kolmayer (1997) confirms the importance of users' expertise in the domain and in the use of information retrieval systems to build successful information access strategies that mix querying and browsing. Kolmayer associates these strategies with different degrees of explicitness of the information problem. Laine-Cruzel and others (1996) built an information retrieval system using plain-text corpus queries that integrate user profiles. These profiles are based on three levels--"cognitive," "technical," and "aim of the search"--that concern the user's domain knowledge; the user's knowledge of plain-text querying tools; and the number, types, and styles of the searched documents.
This prior research helped us build the EKB user profiles out of three axes.
Of course, distinguishing users by their domain expertise levels is one of the essential dimensions of these profiles. Therefore, this characteristic had to be deepened in accordance with the fineness of the knowledge modeling. EKB user profiles identify three user expertise levels in each modeled sub-domain: novice, learner, and expert.
Another dimension of the EKB user profiles concerns users' familiarity with information systems. This dimension mixes the "technical" and "aim of the search" levels described by Laine-Cruzel and her colleagues. It includes the skill levels they define with these systems and their preferred information access strategies. We added to this dimension a list of the users' information research preferences, corresponding to the information system classification criteria (defined later in this article).
These characteristics allow us to define generic user types but not to customize the interface for each individual's preferences. We then added to our profiles a third individual dimension that allows users to specify their viewing preferences and the way they want the EKB to react in navigation or interrogation situations. For example, in this individual dimension, we include stylesheets, specification of information viewing tools, bookmarks that refer either to documents or to diagrams considered by the individual to be particularly useful for his or her usual information needs, and the list of search fields most commonly used on the query construction form.
User profiles act on the EKB interface behavior thanks to a rule-based system (see the section of this article on the EKB design process). Two kinds of rules can be identified. Generic rules, such as application of viewing preferences or information access strategy, can be embodied into a generic EKB architecture. But there also are domain-specific rules, such as the querying form definition, that need to be custom-generated for each project. A series of usability validation tests for the pertinence of these rules must be developed in cooperation with the human factors researchers of the MErLIn project of the French National Institute for Research in Computer Science and Control as soon as the EKB generation workshop is delivered and the first versions of the multi-user EKB are generated.
Modeling the information system to facilitate retrieving information
The purpose of the EKB is to help users find their way in the information system thanks to knowledge modeling. The diagrams present a structuring, task-oriented view of the domain that provides a point of reference for navigating in the domain and accessing the information system. Before the beginning of a project, this system is composed of a disparate and unstructured body of documentary units (that is, the elementary pieces of information of a physical system; it can be a book chapter, an article in a review, or a page of a Web site; but it can also be a single part or the entirety of the collection, in the case of a note or data file). Numerous methods aim to classify such bodies of documents (for example, the Dewey and Classification Decimale Universelle [CDU] systems). Our goal is to elaborate a model that classifies all these units in a suitable manner both for the existing physical classification and for the two other dimensions of the EKB information access strategy.
We want to provide EKB users with "communication points" between this map of the information system and the information system itself. To do so, we need to propose a vision of the information system that matches the domain vision presented by the knowledge diagrams. This is why models of experts' knowledge must be developed at the same time as the inventory of the documents they use to do their work. Restricted by some constraints (detailed in Medini 1999), this approach results in a task-oriented view of the information system.
The ideal way to allow information access from the EKB would be establish the correspondence between each documentary unit and the diagrams describing, for in stance, the tasks that use them or the concepts they refer to. In the case of a very large information system or collection of diagrams, however, this approach would require a considerable amount work that cannot be automated. As a result, we elaborated a domain-specific model that gives a generic description of the information system elements without treating them individually. This model must be customized for each project; however, it is based on a generic reusable structure that can be fleshed out to fit the specific EKB's requirements.
The information system model is composed of a set of criteria that are relevant in regard to physical classification and availability of information, experts' views of the domain, and EKB user profiles. Identifying those criteria is done during the knowledge and user modeling stage. The model is tree-structured, and the identification of the information pieces with these criteria is not unique. Two categories of criteria have been defined.
* The content criteria are related to the domain.
They concern the subdomain, the expertise level, or the activities requiring a specific piece of information. These criteria correspond to both the users' information research preferences, defined above, and the terms of a request, selected from the diagram elements. More precisely, content criteria don't contain all the words in the diagrams, but rather a decomposition of the domain's main classification criteria provided by the experts. Document indexes are built for each criterion, and plain text searches are processed into the corresponding indexes.
* The formal criteria (or document classification criteria) describe the physical form and location of the information system elements. They are composed of classical fields such as title, author, date, type of support, and format, and are used to access the relevant documents or to process direct querying of the information system.
We provide an example of one part of the information system model built for the pilot project later in this article.
The rule-based system associates the diagram elements and the users' characteristics with the values of different criteria. It allows the information retrieval process to be funneled into the corresponding parts of the information system. Technically, the model is implemented as a tree structure of document indexes. These indexes are selected and joined according to the rules. The request is then directed to the junction of these indexes. For users skilled with information retrieval tools, the exact formulation of the request and the indexes list are available. Thus, they can refine the request or directly query the given indexes.
THE EKB DESIGN PROCESS
Because the EKB is supposed to be reusable from one project to another, we present here a design study that resulted in a generic architecture. This architecture will then be customized in each project, using a software tool.
Constraints and functions
Designing an EKB is different from designing regular software. The EKB actually attempts to use previously installed viewers and link them with other tools. The programming task can then be limited to the creation of architecture elements able to
* Define the users' characteristics and information needs
* Store different kinds of information, such as the knowledge diagrams, the information system indexes, and the user profiles
* Display models of the experts' domain knowledge and information use
* Run the rule-based system to dynamically modify the functions and viewing parameters in response to the users' actions, and to call the appropriate tools for information retrieval and viewing
We assumed that a modular architecture would be the most suitable for the EKB. Indeed, because we wanted to be able to automatically generate a specific EKB using a specially designed tool, a modular type of architecture would make such automation easier by separating the different EKB components and identifying the reusable ones and the project-dependent ones. The virtual documents approach that we took provides a technical framework that accommodates all these characteristics.
The design approach: Virtual documents
Virtual documents consist of documents, document parts, or other information pieces that are assembled and formatted according to a given model. They are called virtual because they don't exist as such on a physical device but are generated dynamically. This approach, based on the reusability of data and data structures, also allows information formats and contents to be stored separately. The most advanced technique for generating virtual documents is detailed by Moulin and Pazzaglia (1999): XML/XSL technology is used to compose the document, and then a parser translates it into HTML format viewable with a Web browser.
Generic architecture of an EKB
Figure 4 describes a modular virtual-document-based architecture for the EKB. Its components are
* The databases storing the different types of information (for example, knowledge diagrams, user profiles, stylesheets, interface components, and so forth)
* The different applications needed to display the files
* The EKB core application, which reacts to users' actions and runs the rule-based system to send appropriate requests to the different databases
In Figure 4, the Organization Information database represents the organization's information system and is not included in the EKB architecture. It can be composed of numerous and unhomogeneous information types, so requests are sent to this database using the formal criteria of the information system model. The accessed files may need to be converted into a format that suits the virtual document to be viewed.
A CASE STUDY IN NUCLEAR CRITICALITY SAFETY
To build a prototype EKB, we led a knowledge management pilot project in the Criticality Studies Department (Service d'Etudes de Criticite' or SEC) of the French Nuclear Protection and Safety Institute (Institut de Protection et de Surete Nucleaire or IPSN).
The SEC is the expert body in France on the subject of nuclear criticality safety. Fissile material, such as uranium or plutonium, constantly emits and absorbs neutrons. During chain reactions in nuclear reactors, the ratio between the numbers of emitted and absorbed neutrons equals 1; this is called the criticality point of fissile material. It is vital that when transporting, transforming, or storing these materials, this point is unattainable. The nuclear criticality safety domain consists of being able to define the conditions and constraints that keep each fissile material configuration below this point.
IPSN/SEC is a 50-year-old unit of the CEA that has accumulated an imposing and very diversified information system. Thus, the SEC is an ideal example of an organization where an EKB could help personnel structure and access information. Indeed, for the SEC, there was a real need to organize its information and legacy knowledge to ensure its two main missions: to diffuse its information and knowledge to the French nuclear criticality safety community, and to continue making progress in research in this domain.
In accordance with the generic EKB architecture, three types of models were built. All them were deduced from the modeling stage of the project.
Knowledge models The knowledge management project carried out at IPSN/SEC required 100 hours of interviews of 7 different experts and resulted in about 100 pages of knowledge diagrams, such as the one presented in Figure 5.
This activities diagram shows the lifecycle of a basic nuclear installation (Installation Nucleaire de Base, or INB). Any place where fissile material can be found (that is, where it is created, transformed, or stored)--except nuclear power plants--is an INB. This diagram is shown in French because it is has not been validated in another language.
Six activity subdomains were identified at IPSN/SEC:
* Participation in critical experiences programs
* Development, qualification, and maintenance of criticality safety calculation codes
* Criticality studies for nuclear plant operators
* Analyses of criticality studies for the French nuclear authority
* Definition of French criticality standards and norms
* Diffusion of criticality information and knowledge
These subdomains can be grouped two-by-two, into three subdomain classes:
* Research and development
* French expertise on nuclear criticality safety
The two last subdomains have not been modeled in our pilot project.
User models Identification of the criticality safety subdomains allows us to define the different EKB user types. Of the six identified subdomains, we defined user profiles corresponding only to the four subdomains we modeled, the engineering and the research and development subdomain classes. Table 1 describes the identified user types and their main characteristics. In this demonstration system, we didn't implement the users' individual preference part of the profiles defined earlier in this article. The prototype is currently able to manage only user types instead of real user profiles.
Information models The information system of IPSN/SEC is imposing and heterogeneous. In parallel with our project, IPSN/SEC has begun digitizing its archives. More than 50,000 pages have been digitized and are now available in Adobe[TM] Acrobat PDF format.
We inventoried these information system elements and elaborated a domain-specific model of these system elements in accordance with both their formal description and the way experts use these documents. One part of this model is presented in Figure 6.
The current state of the prototype
The EKB prototype proposes the following functions:
* Navigation in the knowledge diagrams
* Construction of requests
* Processing of a sample of the SEC archive in PDF format
* Sending the requests to a Web search engine
Table 2 briefly describes the functions provided by the prototype to each type of user. Access to applications is accomplished through browser bookmarks.
We are currently integrating into the system a domain-specific search form that will allow the elaboration of complex requests. In future versions, adding XSL stylesheets will allow us to experiment with customizing the look and feel of the interface. Then we will need only to provide users a means to create or choose their own stylesheets and to specify their viewing preferences to propose real user profiles.
An example of the pilot project user interface is presented in Figure 7.
The current EKB prototype user interface is usually divided into two frames. On the left part of the screen, the conceptual map (here, a functional tree) can be selected from a scrolling list. On the right, the information frame can display either a knowledge diagram (if the user is navigating) or a document (if the user is consulting a document).
When navigating, no links are directly available from the diagram nodes. Clicking on a node displays a popup window (on the upper right side of Figure 7) presenting the actions available from this element so the user is aware of the links' destinations. At the bottom of Figure 7, a window shows the results of a previous information search in the PDF document collection.
The final version of the criticality safety EKB is not yet available at IPSN/SEC because we are waiting for the modular EKB generation tool to be delivered. However, the prototype was presented at ICNC '99, the International Conference on Nuclear Criticality-Safety, and IAEA-SR-212/46, the Conference organized by the International Agency for Atomic Energy (see Medini and others 1999 and Chariot 2000). During these demonstrations, potential users seemed very enthusiastic, and it appeared that there was a real need for such tools in many other domains.
In this article, after briefly describing the concept of the electronic knowledge book, we described the conceptual information access strategy that the EKB embodies and the design techniques that can be used to produce it. We then presented a prototype created during an operational pilot project. With this EKB, users can access information either by navigating in a hypermedia presentation of knowledge diagrams or by querying the system directly. They can use the navigation strategy to build a multi-dimensional query that reflects complex information needs. To match the users' presentation preferences, the system integrates user profiles and a customized interface.
Although this prototype doesn't exactly correspond to the announced architecture, it gets closer to this architecture in each new evolution, and the differences often result from technical necessities. Tests must still be conducted with end-users to validate the domain-oriented information access strategy.
The design study and pilot project presented in this article used innovative techniques in different disciplines. All the information about the domain, the users, and the information system were extracted from a global model of the organization's legacy knowledge, obtained using an advanced knowledge management method. This data was then divided among the parts of a modular architecture that can be reused for different projects, using the design approach of virtual documents. In the field of information management, we have purposely developed a generic and task-oriented model of an organization's information system. Our approach incorporates user-centered design because knowledge engineers are necessarily in contact with the intended end-users of the system, and the modeling-validation phase approximates the methods of participatory design.
The prototype shows the technical feasibility of coupling knowledge management and information engineering. Actually, several systems for electronic information management already propose to jointly manage information and knowledge. In these systems, users access knowledge through information. However, with the EKB, users do the opposite: they start navigating in the knowledge model as they would in a map of the information system, and they use this model to access information. A knowledge diagram, then, is meta-information that facilitates information access. Because there are already other conceptual maps to access the knowledge diagrams, we can consider the EKB as a three-layer tool for accessing information, in which an upper layer provides users with a global comprehension of the one below it and facilitates access to it.
LIONEL MEDINI is a PhD student in computer science; he obtained a master's degree in physics from the University of Orsay (France). His doctoral research involves applying human factors techniques to knowledge management methods and information technologies. His research is being conducted in the information technologies department of the French Atomic Energy Commission, where he led a pilot-project and collaborated on the design study for the new version of the Electronic Knowledge Book presented in this article. Contact information: email@example.com
JEAN-MARC CHARLOT studied at the University of Orsay (France) and obtained his PhD in cognitive science from the University of Sherbrooke (Canada). He joined the information technologies department of the French Atomic Energy Commission in 1998. He is mainly interested in computerized applications of cognitive sciences and especially knowledge modeling. He collaborated with C-Log International during the MNESIK project to design an operational framework for generating knowledge and information management tools for the industry. Contact information: firstname.lastname@example.org
MATHIAS CHAILLOT heads the laboratory for information engineering (LII) of the French Atomic Energy Commission's information technologies department. He obtained an engineering degree at ENSIMAG (Grenoble, France) and his PhD in computer science at the University of Grenoble. LII researchers develop methods and tools to improve information production and use by engineers and researchers at the French Atomic Energy Commission. The application domains of LII's activities are knowledge management, scientific intelligence, and technical data management. Contact information: email@example.com
Agosti, M., and A. Smeaton. 1996. "Information retrieval and hypertext." In information retrieval and hypertext, ed. M. Agosti and A. Smeaton, Norwell, MA: Kluwer Academic, pp. 139-178.
Allen, R. B. 1997. "Mental models and user models." In Handbook of human-computer interaction, ed. M. Helander, T. K. Landauer, and P. Prabhu. New York, NY: Elsevier Science, pp. 49-63.
Bachimont, B. 1996. "Du texte a l'hypotexte: Les parcours de la memoire documentaire." Seminaire Interdisciplinaire de sciences Cognitives et Epistemologie Memoires, Inscriptions, Actions, Individuelles et Collectives, 22-26 January, Compiegne, France.
Bates, M. 1989. "The design of browsing and berrypicking techniques for online search." Online review 13, no. 5:405-423.
Belkin, N. J. 1980. "Anomalous states of knowledge as a basis for information retrieval." Canadian journal of information science, 5:133-143.
Bracewell, R., and T. Laferriere. 1996. The contribution of new technologies to (earning and teaching in elementary and secondary schools. Quebec City, PQ: Laval University. http://www.tact.fse.ulaval.ca/fr/html/apport/impact96.html.
Chaillot, M., and J.-L. Ermine. 1997. "Le livre de connaissances electronique," Document numerique 1, no. 1:75-98.
Charlot, J.-M. 2000. "Information engineering based on knowledge modelling: A case study in nuclear criticality safety." International Seminar on Integrated Information Systems, International Atomic Energy Agency, SR-212/46, Vienna, Austria, 10-14 April.
Conklin, J. 1987. "Hypertext: An introduction and survey." IEEE computer (September): 17-41.
Cruz, I. F., and W. T. Lucas. 1998. "Virtual document generation from multimedia repositories." Theory and practice of object systems 4, no. 4:245-260.
Ermine, J.-L, M. Chaillot, P. Bigeon, B. Charreton, and D. Malavieille. 1998. "MKSM: A method for knowledge systems management." In Knowledge management, organization, competence and methodology: Advances in knowledge management, ed. Jos. F. Schreinmakers. Wurzburg, Germany: Ergon Verlag, vol. 1, pp. 288-302.
Kolmayer, E. 1997. "Demarche d'interrogation documentaire et navigation." Preprints of the [4.sup.eme] colloque Hypermedias et Apprentissages, 15-17 October, Poitiers, France, pp. 89-96.
Laine-Cruzel, S., T. Lafouge, J.-P. Lardy, and N. Ben Abdallah. 1996. "Improving information retrieval by combining user profile and document segmentation." In formation processing and management 32, no. 3:305-315.
Leloup, C. 1998. Moteurs d'indexation et de recherche, environnements client-serveur, internet et intranet. Paris, France: Eyrolles.
McNamara, D. S., E. Kintsch, N. B. Songer, and W. Kintsch. 1996. "Are good texts always better?" Learning from texts: Cognition and instruction (January):1-43.
Medini, L. 1997. Conception et realisation d'un livre de connaissances electronique sous forme de document hypermedia. Rapport de projet de DEA, Orsay, et INSTN.
Medini L., J. Anno, M. Chaillot, V. Rouyer, and D.-L. Scapin. 1999. "Towards a knowledge-oriented system for information management in nuclear criticality-safety." Proceedings of the ICNC'99 Conference, September 20-24, Versailles, France, vol. 4, pp. 1451-1561.
Medini, L. 1999. "Utilisation de la modelisation des connaissance pour la conception d'un referentiel hypermedia d'acces l'information scientifique et technique." Proceedings of the Con Conference Interaction Homme-Machine '99, Montpellier, France, 23-26 November, vol. 2, pp. 174-177.
Marchionini G. 1995. Information seeking in electronic environments. New York, NY: Cambridge University Press.
Millerat P., J. L. Ermine, and M. Chaillot. 1996. "Knowledge management for modelling nuclear power plants control in incidental and accidental states. Proceedings of the CESA '96 IMACS Multiconference, Symposium on Modelling, Analysis, and Simulation, Lille, France, vol. 2, pp. 982-987.
Moulin C., and J.-C. Pazzaglia. 1999. Proceedings of the Atelier sur les Documents Virtuels Personnalisables, Con Conferenc IHM '99, Montpellier, France, 23 November, pp. 27-31.
Pamquist, R. A., and Kim Kyung-Sun. 1998. "Modeling the users of information systems: Some theories and methods." In Electronic resources: Use and user behavior, ed. Hemalata Lyer. Binghampton, NY: The Haworth Press, Inc., pp. 3-25.
Ranwez, S., and M. Crampes. 1999. "Conceptual documents and hypertext documents are two different forms of virtual documents." Proceedings of the Workshop on Virtual Documents, Hypertext Functionality and the Web, Eighth International World Wide Web Conference, Toronto, ON, Tuesday May 11, 1999, pp. 21-28. http://www.cs.unibo.it/ [sim]fabio/VD99/ranwez/ranwez.html.
Roger, D., J. Lavandier, and E. Kolmayer. 2000. "Navigation et interfaces: Cartes conceptuelles et autres outils." PARINFO (Programme d'Aide Ia Recherche en Information). http://www.enssib.fr/bibliotheque/documents/ travauz/kolmayer/kolm- navig0.html.
Waem, Y., and S. Hagglund. 1997. "User aspects of knowledge-based systems." In Handbook of human-computer interaction, ed. M. Helander, T.K. Landauer, and P. Prabhu. New York, NY: Elsevier Science, pp. 1159-1176.
TABLE 1: CHARACTERISTICS OF THE IDENTIFIED US TYPES OF THE NUCLEAR CRITICALITY-SAFETY EKB User type Characteristics Engineer High level of knowledge about parameters, expert mechanisms, and tools used for criticality- safety studies and calculations Medium level of knowledge about physical phenomena involved in nuclear criticality safety High level of knowledge about domain-related terminology and most frequently used references "Time pressure" Researcher High level of knowldge about parameters, expert mechanisms, and tools used for criticality calculations High level of knowledge about physical phenomena involved in nuclear criticality safety Knowledge of other related domains (dependent on the expert's background) Atypical use of standard criticality safety tools Learner Medium level of knowledge about parameters, mechanisms, and tools used for criticality calculations Basic level of knowledge about physical phenomena involved in nuclear criticality safety Low level of knowledge about domain-related terminology and of the most frequently used references Low level of knowledge about IPSN/SEC activities Novice No knowledge of the parameters, mechanisms, and tools used for criticality calculations No knowledge of the physical phenomena involved in nuclear criticality safety No knowledge of IPSN/SEC activities TABLE 2: FUNCTIONS OF THE EKB FOR DIFFERENT USER TYPES Sub-domain/User Sub-domain/User Criticality studies Studies-oriented access to (engineering) the whole information system Access to all application Analyses of criticality Expertise-oriented access to studies the whole information (engineering) system Access to all application Calculation codes Codes-oriented access to (research) the whole information system Access to all applications Critical experiences Experiences-oriented access (research) to the whole information system Acess to all application Sub-domain/User Expert Criticality studies Limited access to the (engineering) engineering type information Analyses of criticality Access to the codes studies standard options (engineering) Calculation codes Limited access to the (research) research type information Critical experiences Access to the codes (research) standard options Sub-domain/User Learner Criticality studies (engineering) Analyses of criticality Navigation in the diagrams studies and the closest (engineering) documents through static links Calculation codes (research) Critical experiences (research)
|Printer friendly Cite/link Email Feedback|
|Author:||MEDINI, LIONEL; CHARLOT, JEAN-MARC; CHAILLOT, MATHIAS|
|Date:||Aug 1, 2001|
|Previous Article:||Cruel Pies: The Inhumanity of Technical Illustrations.|
|Next Article:||Plain English? A Study of Plain English Vocabulary and International Audiences.|