Content uploaded by Steffen Staab
Author content
All content in this area was uploaded by Steffen Staab
Content may be subject to copyright.
Knowledge Processes and Ontologies
Steffen Staab , Hans-Peter Schnurr , Rudi Studer , York Sure
Institute AIFB, Universityof Karlsruhe, D-76128 Karlsruhe, Germany
http://www.aifb.uni-karlsruhe.de/WBS
Ontoprise GmbH, Haid-und-Neu Straße 7, 76131 Karlsruhe, Germany
http://www.ontoprise.de
mailto: staab,studer,sure @aifb.uni-karlsruhe.de
mailto: staab,studer,schnurr @ontoprise.de
tel. +49-721-6084751, fax +49-721-6086580
November 29, 2000
Abstract
Technologyfor knowledge managementhas so far focused on the management of knowledge
containers. We present an approach that is oriented towards managing knowledgecontents instead
by identifying knowledge items at various levels of formality. This is done by providing various
types of meta data that are tied to ontologies for conceptual interlinkage. Knowledge items are
embedded into knowledge processes, which are supported by a suite of ontology-basedtools. In
order to handle this sort of rich knowledge process, we introduce a meta process that puts special
emphasis on constructing and maintaining the ontology when introducing knowledge manage-
ment systems. In order to elucidate our approach, we describe a case study about the building of
CHAR, the Corporate History AnalyzeR.
Keywords: Ontology, Knowledge Mangement, Knowledge Process
1 Introduction
In recent years Knowledge Management (KM) has become an important success factor for enter-
prises. Increasing product complexity, globalization, virtual organizations or customer orientation are
developments that ask for a more thorough and systematic management of knowledge — within an
enterprise and between several cooperating enterprises. Obviously, KM is a major issue for human re-
source management, enterprise organization and enterprise culture — nevertheless, information tech-
nology (IT) plays the crucial enabler for many aspects of KM. As a consequence, KM is an inherently
interdisciplinary subject.
IT-supported KM solutions are built around some kind of organizational memory that integrates
informal, semi-formal and formal knowledge in order to facilitate its access, sharing and reuse by
1
members of the organization(s) for solving their individual or collective tasks [2]. In such a context,
knowledge has to be modeled, appropriately structured and interlinked for supporting its flexible
integration and its personalized presentation to the consumer. Ontologies have shown to be the right
answer to these structuring and modeling problems by providing a formal conceptualization of a
particular domain that is shared by a group of people in an organization [5].
There exist various proposals for methodologies that support the systematic introduction of KM
solutions into enterprises. One of the most prominent methodologies is CommonKADS that puts em-
phasis on an early feasibility study as well as on constructing several models that capture different
kinds of knowledge needed for realizing a KM solution [8]. Typically, these methodologies conflate
two processes that should be kept separate in order to achieve a clear identification of issues: whereas
the first process addresses aspects of introducing a new KM solution into an enterprise as well as main-
taining it (the so-called ”Knowledge MetaProcess”), the second process addresses the handling of the
already set-up KM solution (the so-called ”Knowledge Process”) (see Figure 1). E.g. in the approach
described in [7], one can see the mixture of aspects from the different roles that, e.g., “knowledge
identification” and “knowledge creation” play. The knowledge meta process would certainly have its
focus on knowledge identification and the knowledge process would rather stress knowledge creation.
Knowledge Meta
Process
Knowledge
Process
Knowledge Meta
Process
Knowledge
Process
Figure 1: Two orthogonal processes with feedback loops
We start with discussing the interaction between the twotypes ofprocesses in Section 2 of the pa-
per. Here, we contrast the more document-oriented view with the more knowledge item-oriented view
and indicate the implications for the involved knowledge (meta) processes. The notion of knowledge
items is further elaborated in Section 3. Special emphasis is put on the various formality levels of
these knowledge items and the role meta data play in this context. In Section 4, we elaborate in detail
the second type of process, the Knowledge Process. We set up an ontology-based approach and iden-
2
tify five subprocesses that constitute the Knowledge Process: knowledge creation, knowledge import,
knowledge capture, knowledge retrieval and access, and knowledge use. As we will see, ontologies
constitute the glue to tie together all these knowledge subprocesses. Furthermore, ontologies open the
way to move on from a document-oriented view on KM to a contents-oriented view of KM, where
knowledge items are interlinked, combined and used within inferencing processes.
In section 5 we define a methodology for introducing an ontology-based KM solution into enter-
prises, i.e. we address the first type of process mentioned above, the Knowledge Meta Process. The
methodology extends and improves the CommonKADS methodology [8] by introducing — among
others — specific guidelines for developing and maintaining the respective ontology. Special empha-
sis is put on a stepwise construction and evaluation of the ontology.
We illustrate our concepts and methods by outlining the development and usage process of a
concrete KM solution, i.e. the Corporate History AnalyzeR (CHAR) (cf. Section 6). CHAR is a
valuable example for the gain in functionality and usability that results from our ontology-based KM
approach. Flexible view generation, integrated results and knowledge inferred as part of KM queries
constitute functionalities that can be achieved by such an integrated approach. Our approach shows
that one can achieve a clear identification and handling of the different subprocesses that drive the
development and usage of KM applications. All these subprocesses are supported by appropriate
tools that are tied together by the ontology infrastructure (see for more details [9]).
Knowledge Management is a process which is not only governed by IT. Hence, one needs to keep
the balance between human problem solving and automated IT solutions. This balancing distinguishes
KM from traditional knowledge-based systems. Nevertheless, the extensive knowledge modeling
tasks that are inherent in ontology-based KM approaches support Alun Preece’s saying “Every KM
project needs a knowledge engineer”.
2 Knowledge Items in Knowledge Processes and Meta Processes
The core concern of IT-supported knowledge management is the computer-assisted capitalization
of knowledge [1]. Because information technology may only deal with digital, preferably highly-
structured, knowledge the typical KM approach distinguishes between computer-based encoding in
an organizational memory and direct transfer that is done by humans. Sticking to what is almost
readily available, KM systems have tended to serve either the needs of easy access to documents
3
Table 1: Approaching the Knowledge Process — two extreme positions
Build an
Infrastructure
for your organizational memory system
Re-organize
processes
to deal with creation and distribution of
knowledge
4.
Organize the
knowledge processes
to allow for creation, handling, and process
support of and around knowledge items
Build an
Infrastructure
for your organizational memory system
3.
Find out which
knowledge items
deal with these knowledge needs
Find out which
business documents and databases
deal with these knowledge needs
2.
Find out what the core knowledge needs are1.
Knowledge item focusDocument focus
Build an
Infrastructure
for your organizational memory system
Re-organize
processes
to deal with creation and distribution of
knowledge
4.
Organize the
knowledge processes
to allow for creation, handling, and process
support of and around knowledge items
Build an
Infrastructure
for your organizational memory system
3.
Find out which
knowledge items
deal with these knowledge needs
Find out which
business documents and databases
deal with these knowledge needs
2.
Find out what the core knowledge needs are1.
Knowledge item focusDocument focus
(e.g., building on groupware, etc.) or the encoding of knowledge that facilitates the direct transfer of
knowledge by humans (e.g., by people yellow pages, skill databases, etc.).
Introducing KM to a company (i.e., moving along the Knowledge Meta Process in “the yellow
circle” in Figure 1), a very simple, pragmatic approach has typically been pursued, which however
meant that only the low hanging fruits were picked. This approach is summarized in the left column of
Table 1. What appears preemminent in this approach is the focus on the handling of documents (steps
2 and 3) and the existing, but minor role of the appendix “process”. In spite of its immediate successes,
this approach shows several disadvantages. In particular, it often leads to the consequence that the
Knowledge Process steps (“the blue circle”) of creation, import, capturing, retrieving/accessing, and
using are only very loosely connected, if at all (cf. Figure 2). The underlying reason is that for each of
these steps different types of business documents play a major role, which makes “knowledge re-use”
— and not only knowledge re-finding — extremely difficult.
Focus on the Knowledge Process. In fact, there are several commercially sucessful solutions that
focus on the process thought of handling knowledge, but due to the evolving nature of business so-
lutions they have not quite reached the level that allows for the immediate connection of knowledge
with process steps and the overarching combination of knowledge items across several process steps.
4
From knowledge containers to knowledge items. Subsequently, we show how domain ontologies
may act as the glue between knowledge items, bridging between different Knowledge Process steps.
Thereby, we argue for a refocus on the Knowledge Process and its core items, which need not be
documents! This shift becomes visible in the second column of Table 1, which positions knowledge
items and Knowledge Processes in the center of consideration.
The reader may note that we contrast two rather extreme positions in Table 1. As becomes obvious
in recent research papers, current knowledge management research tends to move away from the
document focus to a focus on knowledge items and processes [1, 10]. While for a multitude of
settings we still see the necessity for the document-oriented view, we argue for a more constructivist
view of the Knowledge Processes. In particular, we believe that the exuberant exchange and trading
of knowledge within and across organizations still has to begin — and that it needs a knowledge
item-oriented view such as we plead for.
3 Knowledge Items
Relevant knowledge items appear in a multitude of different document formats: text documents,
spreadsheets, presentation slides, database entries, web pages, construction drawings, or e-mail, to
name but a few. The challenge that one must cope with lies in the appropriate digestion of the
knowledge, e.g., by “simple” reuse, or by aggregation, combination, condensation, abstraction, and
by derivation of new knowledge from aggregations. Following only the lines of traditional document
management, IT support for knowledge management cannot take advantage of the contents of the
business documents, but only of its explicit or implicit classification. At the other extreme of this
spectrum, there are expert systems that structure and codify all the knowledge that is in the system.
Though such an approach may sometimes be appropriate, it is certainly not the way to follow in the
typical knowledge management scenario, where not everything can be codified, a lot of knowledge is
created sporadically, and the worth of knowledge re-use is only shown over time and not necessarily
obvious from the very beginning.
Hence, one must search for the adequate balance between reuse, level of formality, and costs
to codify knowledge. For instance, certain helpdesk scenarios imply long term use of extremely
well-defined knowledge items [4]. Then it may be worth to codify extensively and to spend some
considerable amount of time and money on coding. On the other hand, a sporadic discussion is
5
typically not worth coding at all, since it lives on the spur of the moment and often is negligible and,
hence, not reusable after some short time.
As a way to balance these conflicting needs and to flexibly manage various degrees of encoded
knowledge, we advertise the use of various notions of meta data. The different notions of the term
“meta data”, i.e. data about data, may be classified at least into the following categories:
1. Data describing other data. We may again divide this category into twoorthogonal dimensions.
(a) The one dimension concerns the formality of this data. Meta data may range from very
informal descriptions of documents, e.g. free text summaries of books, up to very formal
descriptions, such as ontology-based annotation of document contents.
(b) The second dimension concerns the containment of the meta data. Parts of meta data may
be internal to the data that is described, e.g. the author tag inside of HTML documents,
while others may be stored completely independently from the document they describe,
such as a bibliography database that classifies the documents it refers to, but does not
contain them.
2. The second major connotation of meta data is data that describes the structure of data. For our
purpose, one might refer to this notion by the term “meta meta data”, because we describe the
structure of meta data. Also, in our context this notion boils down to an ontology that formally
describes the domain of the KM application, possibly including parts of the organization and
the information structures (cf. [1]). The ontology allows to combine meta data from different
parts of the Knowledge Process and data proper that adhere to the ontology description.
Meta data in its first connotation fulfills a double purpose. It condenses and codifies knowledge
for reuse in other steps of the KM process by being connected through mutual relationships and
the ontology (the meta meta data). Furthermore, it may link knowledge items of various degrees of
formality together, thus allowing a sliding balance between depth of coding and costs.
4 Knowledge Process
Once a KM system is fully implemented in an organization, knowledge processes essentially circle
around the following steps (cf. Figure 2).
6
Knowledge creation and/or import,i.e. contents need to be created or converted such that they
fit the conventions of the company, e.g. to the the knowledge management infrastructure of the
organization;
then knowledge items have to be captured in order to elucidate importance or interlinkage, e.g.
the linkage to conventionalized vocabulary of the company;
retrieval of and access to knowledge satisfies the “simple” requests for knowledge by the knowl-
edge worker;
typically, however, the knowledge worker will not only recall knowledge items, but she will
process it for further use in her context.
Figure 2: The Knowledge Process
Knowledge Creation. Creation of computer-accessible knowledge proceeds between the two ex-
tremes of very formal and very informal knowledge. What is often overlooked is that comparatively
7
deep coding can often be done without requiring any extra efforts. Business documents, in general, are
not arbitrarily changing knowledge containers, but reasonably often come with some inherent struc-
ture, part of which is often required by quality management and, e.g., engineering requirements. Thus,
in a contribution to [10] we have proposed to embed the structure of knowledge items into document
templates, which are then filled on the fly by doing daily work. The granularity of this knowledge
then lies in the middle between the extremes of coarse representations of business documents only
and an — for the purpose of KM — overly fine one, such as found in expert systems. Thus, one finds
several degrees of formality between the two extremes of very formal and very informal knowledge.
We compare some of them in Table 2.
Table 2: Degrees of formal and informal knowledge
Degree Model Interface Example
Thoroughly Formal Relational Form Interface Database interface
Formal Content-structured
document Tight XML Structure XML-EDI
Partially Formal Document template Loose XML Structure Investment recom-
mendation template
(cf. Table 3)
Informal Free text No predefined structure ASCII text file
In this comparison, we use the term “content-structured documents” to refer to XML structures
that are tightly (sometimes explicitly, sometimes implicitly) linked to a domain model. For instance,
XML-EDI documents come with a predefined structure alluding to a standard framework for exchang-
ing data, such as invoices, healthcare claims, or project status reports. By document templates we refer
to similar structures, which however come with a larger degree of freedom, including large chunks of
informal knowledge items (e.g.,cf. Table 3). One may note that these different degrees of formality
are often combined, e.g. unstructured documents may have attached a form for adding Dublin Core
meta data.
Careful analysis of the knowledge items in use allows for the possibility to add formal knowl-
edge parts into the process of creating these documents, thus pushing the degree of formality slightly
upwards without endangering the overall usage of the system, which could be incurred by an expert
systems-like approach to Knowledge Management.
8
Table 3: Filled Investment Recommendation Template.
investmentrecommendation
author Henrik Oppermann /author
plandate October 18th, 2003 /plandate
interviewpartners name York Sure /name
name Hans-Peter Schnurr /name
name Steffen Staab /name
/interviewpartners
recommend strong buy /recommend
details peergroup /peergroup
/details
/investmentrecommendation
Knowledge Import. For many KM purposes the import of knowledge items into the KM system
of the organization has the same or more importance than their creation within the organization. The
overall situation is akin to data warehousing — only that the input structures are more varying and
the target structures are much richer and more complex than this is the case for the standard data
warehouse.
For imported knowledge accurate access to relevant items plays an even more important role than
for “home-made” knowledge. The reason is that for home-made knowledge items, people may act as
a backup index. This is not the case for recently imported knowledge that no one has seen yet. In
fact, access studies to KM systems have shown that OM parts that cover imported knowledge are less
heavily exploited than those covering home-grown ones [6] — though it seems implausible that they
would contain less useful contents.
Knowledge Capture. Once that knowledge items have been created, but not yet, or only incom-
pletely, captured apart from their context, e.g. from their database entries or their business document
containers, the next process step is the capturing of their essential contents. Besides of common in-
dexing and abstracting techniques known from library science, we provide meansto capture document
excerpts as well as interlinkage between excerpts by our tool, OntoAnnotate (cf. Figure 3).
OntoAnnotate allows to create objects and describe them with their attributes and relations exploit-
ing knowledge that is found on web pages, in spreadsheets, or in text documents. Thus, in the example
we capture the facts that the company M.A. Hanna sells to GE its Shape Distribution Business, link
them to the excerpts from which they origin and note who has done the annotation.
9
Figure 3: Capturing Knowledge about a Selling
By this annotation process meta data are created that conform to the ontology and, hence, can be
aligned with related information to yield analyses and derivations such as described in our case study
in Section 6. The origins of the meta data may be used to validate the overall information.
Knowledge Retrieval and Access. Large parts of knowledge retrieval and access from an ontology-
based organizational memory are performed through conventional GUI, exploiting means like infor-
mation retrieval or taxonomy-enhanced database views. In addition, one may use the ontology to
derive further views. In particular, we exploit the ontology for navigation purposes (like Yahoo).
Thus, knowledge workers may explore what is in the organizational memory without being required
to ask a particular question — which is often a hard task for newbies. Also, the ontology allows to
derive additional links and descriptions, e.g. the ontology allows to derive state descriptions for points
in time for which no explicit data exists, or it provides new hyperlinks that are not given explicitly.
Thus, we may complete views without requiring that all information is given. An actual example for
the latter is given in detail in Section 6.
10
Knowledge Use. Knowledge use deals with the most intricate points of knowledge management. It
is the part that is most often neglected, because many KM systems assume that once some relevant
document is found everything is done. Eventually, however, the way that knowledge from the orga-
nizational memory is used is quite involved. Therefore topics like proactive access, personalization,
and, in particular, tight integration with subsequent applications play a crucial role for the effective
re-use of knowledge. Very often it is not even the knowledge itself which is of most interest, but the
derivations that can be made from the knowledge. For instance, in our case study we will see that no
single knowledge items about a company may be relevant to a market analyst, but the overall picture
presented by analysis.
In addition, usage data tells a lot about the organizational memory and about the organization. For
instance, one may analyze which processes, customers and techniques are tied to core figures of the
organization.
5 Knowledge Meta Process: Methodology for Ontology-based Knowl-
edge Management
Ontologies aim at capturing domain knowledge in a generic way and provide a commonly agreed un-
derstanding of a domain, which may be reused and shared across applications and groups. Ontologies
typically consist of definitions of concepts, relations and axioms. Until a few years ago the building
of ontologies was done in a rather ad hoc fashion. Meanwhile there have been some few, but seminal
proposals for guiding the ontology development process (e.g. [11]). For instance Guarino & Welty [3]
give formal guidelines for constructing a consistent and reusable ontology.
In contrast to these methodologies, which mostly restrict their attention within the ontology itself,
our approach focuses on the application-driven development of ontologies. We cover aspects from
the early stages of setting up a KM project to the final roll out of the ontology-based KM application.
Here, we focus our attention on the development of the ontology. Thereby, we integrate some lessons
learned of our practical experiences into the steps to be taken to perform each ontology development
activity. The ontology development process is sketched in Figure 4, we will now describe its steps in
detail.
11
Figure 4: The Knowledge Meta Process
Feasibility study. Any knowledge management system may only function satisfactorily if it is prop-
erly integrated into the organization in which it is operational. Many factors other than technology
determine success or failure of such a system. To analyze these factors, we must initially perform
a feasibility study to first identify problem/opportunity areas and potential solutions, and second, to
put them into a wider organizational perspective. The feasibility study serves as a decision support
for economical and technical project feasibility, in order to select the most promising focus area and
target solution. An approach for carrying out a feasibility study is described by the CommonKADS
methodology (cf. [8]). It should be carried out before actually developing ontologies and serves as a
basis for the kickoff phase.
Kickoff phase for Ontology Development. The output product of the kickoff phase is an ontology
requirements specification document (cf. Table 4) describing what an ontology should support and
sketching the planned area of the ontology application. It should also guide an ontology engineer to
decide about inclusion, exclusion and the hierarchical structure of concepts in the ontology. In this
early stage one should look for already developed and potentially reusable ontologies. In summary, it
12
should clearly describe the following information:
Table 4: Ontology Requirements Specification Document (ORSD)
1. Goal of the ontology
2. Domain and scope
3. Applications supported by the ontology
4. Knowledge sources (e.g. domain experts, organization charts, business plans, dictionairies,
index lists, db-schemas etc.)
5. Potential users and usage scenarios
13
6. Competency questionnaire (i.e. an overview of possible queries to the system, indicating the
scope and content of the domain ontology, cf. Table 5)
7. Potentially reusable ontologies
Table 5: Competency Questionnaire
Refinementphase. The goal of the refinement phase is to produce a mature and application-oriented
target ontology according to the specification given by the kickoff phase. This phase is divided into
different subphases:
1. The gathering of an informal “baseline taxonomy” containing relevant concepts given during
the kickoff phase.
14
2. A knowledge elicitation process with domain experts based on the initial input from the baseline
taxonomy to develop a “seed ontology” containing relevant concepts, relations between them
and axioms on top. The seed ontology is usually expressed at an epistemological level.
3. A conceptualization and formalization phase to transfer the seed ontology into the “target on-
tology” expressed in formal representation languages like F-Logic, OIL or DAML-ONT.
The usage of potentially reusable ontologies (identified during the kickoff phase) may improve
the speed and quality of the development during the whole refinement phase. These ontologies might
e.g. give useful hints for modeling decisions.
Evaluation phase. The evaluation phase serves as a proof for the usefulness of developed ontologies
and their associated software environment. In a first step, the ontology engineer checks, whether the
target ontology suffices the ontology requirements specification document and whether the ontology
supports or “answers” the competency questions analyzed in the kickoff phase of the project. In a
second step, the ontology is tested in the target application environment. Feedback from beta users
may be a valuable input for further refinement of the ontology. A valuable input may be as well the
usage patterns of the ontology. The prototype system has to track the ways users navigate or search
for concepts and relations. With such an “ontology log file analysis” we may trace what areas of
the ontology are often “used” and others which were not navigated. This phase is closely linked to
the refinement phase and an ontology engineer may need to perform several cycles until the target
ontology reaches the envisaged level — the roll out of the target ontology finishes the evaluation
phase.
Maintenance phase. In the real world things are changing — and so do the specifications for on-
tologies. To reflect these changes ontologies have to be maintained frequently like other parts of
software, too. We stress that the maintenance of ontologies is primarily an organizational process.
There must be strict rules for the update-delete-insert processes within ontologies. We recommend
that the ontology engineer gathers changes to the ontology and initiates the switch-over to a new
version of the ontology after thoroughly testing possible effects to the application, viz. performing
additional cyclic refinement and evaluation phases. Similar to the refinement phase, feedback from
users may be a valuable input for identifying the changes needed. Maintenance should accompany
ontologies as long as they are on duty.
15
In the following section we bring this general methodology to life in our case study CHAR.
6 Case Study: the Corporate History AnalyseR — CHAR
6.1 Knowledge Meta Process instantiated: The development of CHAR
Feasibility study. The active tracking and management of knowledge relevant to one’s business is
a major task for knowledge-intensive companies. While the correct analysis of market situations and
competitors are critical requirements for success, the failure to provide adequate knowledge about
one’s business environment may incur large losses. The trouble is that management and professionals
have a hard time gathering information, analyzing it, and performing their operational work. Here
comes corporate research into play. The task of the corporate researcher is the tracking of relevant
knowledge in the outside business setting and the communication of important knowledge to stake-
holders within the company. Nowadays, the market analysts, consultants and inhouse market research
departments try to track the activities of their industry with traditional methods. Newspaper articles,
online databases and annual company reports are analyzed and competitors web pages are thoroughly
tracked. The results are presented to the management, published in reports and distributed. There
exist several problems in this research process:
Information archives are document-based. For a collective gathering of facts, this view is too
coarse.
Typical document management systems rely almost exclusively on information retrieval tech-
niques, which are too inaccurate given the task at hand.
Implications may only be made transparent if background knowledge is used. E.g., if one com-
pany sells one of its units, the implication is that this unit may no longer be the sole supplier for
the company as it was before. Nowadays, systems are typically not supporting such background
knowledge.
Different people may contribute similar knowledge, but they may need different views onto the
same basic piece of information.
A KM system that covers such knowledge about the outside world should, (i), support the collec-
tive gathering of information on the level of facts rather than documents, (ii), integrate the gathering
16
task smoothly into the common research process, (iii), allow to intelligently combine facts, (iv), check
new facts against the available background knowledge, (v), allow a multiple view access to the knowl-
edge via a single entry portal, and, (vi), allow to route derived facts back into the common workplace
environment. With these aims in mind, we have developed the ontology-based application CHAR, the
Corporate History AnalyzerR.
Kickoff phase for Ontology development. Starting from these requirements, the question was how
to bring the required conceptual structures and reasoning capabilities into action following the further
steps of our Knowledge Meta Process shown in Figure 4. User requirements were collected in the
requirements specification phase during structured interviews with corporate research analysts. The
Ontology Requirements Specifications Document shown in Table 4 has been an output of this first
phase. There, the ontology had to be specified as described in Section 5. In a next step, the ontology
had to be developed in detail. Therefore, the domain experts were asked, what questions they would
expect to be answered by a system supporting their corporate research work. These competency
questions allowed to find the most important concepts and relations between concepts. One result
of analyzing the questionnaires has been that the system should deliver answers about the acquisi-
tions, mergers and restructuring of companies over specific time periods. The example Competency
Questionaire in Table 5 lists competency questions CQ1 to CQ5 that were found during the ontology
kickoff. In addition, the need for a clarification of organizational “wording” was recognized, e.g. “Is
a business unit the same as a division or a department?”
Refinement phase. First in the refinement phase, the concepts were brought into a taxonomic hier-
archy by the knowledge engineer. Figure 5 shows some part of the taxonomic structure.
Second, the taxonomy was shown to the domain experts to gather additional concepts and relevant
attributes and relations between the concepts. This work was supported by our ontology engineering
tool OntoEdit (cf. [9]), which allows the epistemological modeling of the ontology and the formaliza-
tion in different representation languages, like e.g. F-Logic, OIL and DAML-ONT. One particularity
of the corporate history domain was that time had to be modeled. There are many objects which have
a life time. For instance a company has a start time when it is founded and there is an end time when
this company is acquired by another company, is merged with another company, or goes bankrupt.
So duration had to be modeled. Likewise, an “Acquisition” is an “Event” which occurs at a specific
point in time, requiring a model for time points. One of the key characteristics of CHAR is that the
17
Figure 5: CHAR taxonomy
user provides facts to the system about actions like acquisitions and mergers and the system inferes
the consequences. For this purpose rules had to be modeled for all possible activities which incur
such consequences. For instance, “If two companies are merged, a new company with a new name is
created, the old companies ceases to exist, they are from now on subsidiaries of the new company” or
18
Figure 6: CHAR
“If a division is outsourced it becomes part of a new company or it becomes a company on its own.”
For our concrete example, CHAR, a Web interface had to be developed that used the ontology
for querying and augmenting the knowledge repository (cf. Figure 6). There, we needed to perform
a query development step to formalize the views and competency questions described earlier. This
development step depended mostly on the actual application setting. It was independent from the
ontology construction, on which the paper reports.
Evaluation phase. In the evaluation and testing phase the usability of the ontology was investi-
gated. With regard to the CHAR ontology we found that the competency questions CQ2 to CQ5
(Table 5) could be handled successfully by the ontology. However, it had been envisaged that com-
plex knowledge content could be queried and depicted on one screen, such as company purchases,
19
regions of previous activities and regions of activities of the purchased company. Beta testing the
system the research analyst found that singleton knowledge items were spread over different screens
of the application. Thus, they were very hard to combine by the research analyst in order to answer
strategic questions. Hence, he wanted better support from the ontology. Through the discussion we
came up with additional competency questions (CQ6 and following) that should be handled by the
ontology. Hence, corresponding axioms were added to the ontology and a new query window for
strategic questions was introduced to the application.
Maintenance phase. We are currently in a maintenance phase, where through a shift in the goal ori-
entation of the application, we face requirements for considerable extension of the ontology and the
application. Besides of corporate histories, the system is now required to support the extended track-
ing of market circumstances, including, e.g., a more detailed view of available technology and better
comparisons with peer groups. Major problems that we now must face are, first, all the documenta-
tion and versioning problems that have so far been neglected in practically all ontology development
environments. Also, there are ontology parts for which our development methodology had not yet
been followed. There, developers who were new to the project often need to reverse-engineer some
ontology structures in order to find answers about their actual usage. For the future, we expect that
comprehensive use of the proposed methodology will mostly prevent these problems.
6.2 Knowledge Process instantiated: Usage of CHAR
CHAR, the Corporate History AnalyzeR should allow many people to contribute factual knowledge
in a way that is embedded into their common work process and that is organized around a semantic
background. On the other hand, to deliver answers, it should provide multiple views onto the same
knowledge base for different time frames, for different regional foci, for varying intra-organizational
structures and for different strategic questions, to name but a few.
Providing Knowledge. The process of providing new facts into the knowledge warehouse should
be as easy and as smoothly integrated into the common working tasks as possible. For this reason
we offer various modes of contributing knowledge. First, one may enter information through a form-
based interface. Second, when the information that is to be provided is produced during the writing
of documentations or reports one may use a template approach in order to generate knowledge by
writing these documents. Third, one may use wrapper mechanisms in order to provide data from
20
01.10.93
01.04.97
Figure 7: Comparing two Organizational Structures
tables and lists on the web. Fourth – and most important for CHAR, one may use our annotation tool
in order to add meta data to data given in documents. A snapshot of the annotation tool noting some
action about M.A.Hanna is shown in Figure 3: The user reads or works with documents using a text
or spreadsheet processing tool or an internet browser. When he detects some relevant change being
described in the document and this change might become relevant, he highlights the word or phrase
and uses the annotation tool to select the type of the phrase (e.g. “M.A.Hanna is a company”) and its
relation to other material (e.g. “Hanna sells Shapes Distribution Business to GE Plastics on May 11,
2000”). The document, these facts and meta data about the annotater, the time of annotation, etc. is
all stored in the back-end knowledge repository.
21
Querying for Knowledge. The query interface of the Corporate History Analyzer has been de-
veloped in order to deal with organizational and strategic questions that depend on spatio-temporal
constraints. It renders views that may be seen on a common web browser. Actually, they look just like
common web sites with some choices for selection. These selections are also controlled by the knowl-
edge of the backend system, e.g. one may select companies that are known to exist in the knowledge
repository. Figure 6 depicts the main views that are offered by CHAR, viz. Activities, Organization,
Know How, Strategic Questions and General Query Possibilities (indicated by “Search”). The first
major category of queries relevant to the corporate history is about organizational structures and the ac-
tivities that change organizational structures. For instance, the view of “Acquisitions of M.A.Hanna”
returns all its purchases and corresponding views are offered for Selling, Mergers, Restructuring and
Management Changes (Figure 7). What is interesting to note at this point is that it is rather difficult
to get a clear picture of what is really happening with M.A.Hanna. It is difficult and time-consuming
for the human analyst to detect some trend in lists of single knowledge items. Observations become
much easier, when different types of knowledge items may be related and contrasted. For instance,
Figure 7 depicts two snapshots of the organization of M.A.Hanna that are automatically derived from
single activities, like acquisitions and restructurings, and that give the analyst a neat picture of how
formerly isolated purchases that M.A.Hanna made before 1994 were more tightly integrated in the
company in 1997 (e.g. “Compounding Technology” having been reorganized into the Business Area
“Plastic Compounding”).
Thus, besides sophisticated support that is based on concrete facts and figures and that does not
aim at interpretation thereof, CHARsupports strategic questions indicating possible answers to ques-
tions about business competitors that cannot be answered definitely, but that rely on some conjectures.
For instance, the purchase of a company from abroad may lead to a gain of market share in that area,
and thus to a regional diversification.
7 Conclusion
The importance of knowledge processes has been rapidly recognized lately. We here have shown
that there are (at least) two dimensions on which these Knowledge Processes need to be analyzed:
whereas the first dimension considers the Knowledge Meta Process that is needed to introduce a
knowledge management solution into an enterprise, the second dimension addresses the Knowledge
22
Processes that are performed while running a knowledge management solution. During the Knowl-
edge Meta Process an ontology is developed that acts as the conceptual glue between different steps of
the Knowledge Process proper. The development of the Corporate History Analyzer is an intriguing
example that highlights the assignments of the different tasks to steps in our overall framework.
Of course, the conceptualization of domains and the related meta data will have to reflect the
changing environment of the enterprise. I.e. evolving ontologies and, thus, evolving metadata will
become an important aspect for maintaining a knowledge management solution. For the future, we
therefore envision that the (semi-)automatic analysis of the Knowledge Processes feeds back into the
Knowledge Meta Process cycle. For instance, new concepts that come up during the use of the KM
solution may be introduced into the ontology — which is maintained following the methodology for
KM system development and maintenance. The challenges then lie, (i), in ontology learning and
evolution for analysing KM processes and, (ii), in methodological issues for dynamically updating
the KM solution. Our framework may then leverage these evolving systems, because it offers a clear
distinction of relevant processes and responsibilities.
Acknowledgements. We thank our colleagues and students at AIFB, University of Karlsruhe, and
ontoprise GmbH, J. Angele, S. Decker, M. Erdmann, A. Hotho, A. Maedche, M. Maier-Collin, D.
Oberle, D. Wenke, without whom the research would not have been possible. The research presented
in this paper has been partially funded by EU in the IST-1999-10132 project On-To-Knowledge and
by BMBF in the project GETESS (01IN901C0).
References
[1] A. Abecker, A. Bernardi, K. Hinkelmann, O. K¨uhn, and M. Sintek. Toward a technology for
organizational memories. IEEE Intelligent Systems, pages 40–48, May / June 1998.
[2] R. Dieng, O. Corby, A. Giboin, and M. Ribiere. Methods and tools for corporate knowledge
management. Int. Journal of Human-Computer Studies, 51(3):567–598, 1999.
[3] N. Guarino and C. Welty. Identity, unity, and individuality: Towards a formal toolkit
for ontological analysis. Proceedings of ECAI-2000, August 2000. available from
http://www.ladseb.pd.cnr.it/infor/ontology/Papers/OntologyPapers.html.
[4] L. Morgenstern. Inheritance comes of age: Applying nonmonotonic techniques to problems in
industry. Artificial Intelligence, 103:1–34, 1998.
[5] D. O’Leary. Using AI in knowledge management: Knowledge bases and ontologies. IEEE
Intelligent Systems, 13(3):34–39, May/June 1998.
[6] D. O’Leary. Knowledge management: An analysis based of knowledge use and reuse. IEEE
Intelligent Systems, 2001.
23
[7] G. Probst, K. Romhardt, and S. Raub. Managing Knowledge. J. Wiley and Sons, 1999.
[8] G. Schreiber, H. Akkermans, A. Anjewierden, R. de Hoog, N. Shadbolt, W. Van de Velde, and
B. Wielinga. Knowledge Engineering and Management — The CommonKADS Methodology.
The MIT Press, Cambridge, Massachusetts; London, England, 1999.
[9] S. Staab, J. Angele, S. Decker, M. Erdmann, A. Hotho, A. Maedche, H.-P. Schnurr, R. Studer,
and Y. Sure. Semantic community web portals. In WWW9 — Proceedings of the 9th International
World Wide Web Conference, Amsterdam, The Netherlands, May, 15-19, 2000. Elsevier, 2000.
[10] S. Staab and D. O’Leary, editors. Bringing Knowledge to Business Processes. Papers from 2000
AAAI Spring Symposium, Technical Report SS-00-03, Menlo Park, CA, 2000. AAAIPress.
[11] M. Uschold and M. Gruninger. Ontologies: Principles, methods and applications. Knowledge
Sharing and Review, 11(2), June 1996.
24