SPINNING A CORPORATE SEMANTIC WEB FOR

PRODUCT ENGINEERING

Uwe Keller, Stijn Heymans

Digitial Enterprise Research Institute (DERI), University of Innsbruck, Technikerstrasse 21a, Innsbruck, Austria

Alois Reitbauer

ProFactor Produktionsforschungs GmbH, Im Stadtgut A2, Steyr-Gleink, Austria

Michael Neswal

ATENSOR Engineering and Technology Systems GmbH & CoKG, Im Stadtgut A2, Steyr-Gleink, Austria

Keywords:

Knowledge Management, Semantic Web, Multi-Agent System, Collaborative Work, User-centered Support.

Abstract:

We propose a novel approach towards a generic engineering support environment that combines Semantic

Web technologies, Semantic Desktops, Group Memory Systems and Multi-Agent Systems to overcome the

problems of current systems for supporting individuals in collaborative engineering processes. In particular,

we aim at transferring the principles underlying the Semantic Web into an enterprise. This paper provides a

motivation for our research, identiﬁes requirements and outlines the proposed solution.

1 INTRODUCTION

Engineering of products is a complex and ubiquitous

task; it occurs within enterprises across all indus-

tries and faces products of increasing complexity. As

shown in Fig. 1, already for the process of design-

ing a single product, involved artifacts are numerous

and highly inter-dependent. Information about de-

sign elements (or parts) is created and managed by

different people using their favorite tool. People can

take very different roles in the process (e.g. mechan-

ical engineer, software developer, product designer,

project manager, ﬁnancial controller etc.) and there-

fore generate information about parts of the product

that reﬂect their speciﬁc roles. The result is an ex-

tensive use of various data formats to represent cer-

tain information about the single parts of the product.

At the same time, there is neither a single uniform

data format that could suitably serve all the people

involved in the process, nor is it possible to impose

the use of such a uniform, generic format since the

used tools are not natively designed to serve an inte-

grated, process-wide environment. Hence, traditional

database systems are not applicable for data manage-

ment in this setting (Franklin et al., 2005).

Design decisions for certain parts of a product

must be based on all relevant information. This in-

Figure 1: A Typical Scenario when Designing a Product.

cludes information about all other parts that are in

some way directly or indirectly related to the ele-

ment under consideration. A decision on how an el-

ement has to be designed might affect (or be con-

strained by) all these other parts: for instance, if

there is a need to change the diameter and material

of a screw for cost reasons, all regions around the re-

spective drill wholes where this screw is supposed to

be used as well as all physical connections to other

parts of the product have to be checked to see if there

would be any subsequent problems (e.g. geometric,

mechanical, physical ones etc.) with the new de-

169

Keller U., Heymans S., Reitbauer A. and Neswal M. (2007).

SPINNING A CORPORATE SEMANTIC WEB FOR PRODUCT ENGINEERING.

In Proceedings of the Third International Conference on Web Information Systems and Technologies - Society, e-Business and e-Government /

e-Learning, pages 169-174

DOI: 10.5220/0001287301690174

 SciTePress

sign. Design in this respect can be understood as

high-dimensional multi-parameter optimization prob-

lem with constraints stemming from different areas

(such as mechanical or or geometric restrictions). The

product design team altogether tries to ﬁnd an optimal

solution to the given constraints. Search in the solu-

tion space is knowledge-driven and requires intensive

communication and collaboration:

Information in general is distributed across differ-

ent legacy systems that are rarely integrated or even

inter-operable. Therefore, it is not easy for individ-

uals to identify and to get needed information for a

speciﬁc task. It is often necessary to talk to a series of

people to eventually ﬁnd out the desired information.

Furthermore, design decisions can often not be taken

by individuals alone, but must be taken by a group

of experts. Typically, those people are not located at

the same site, but work at different places, even in

different time zones. Face-to-face communication is

therefore not always possible and time-lags to iden-

tify relevant information or to discuss design issues

for resolving a concrete problem may arise.

In fact, the overall situation is even more com-

plicated in a real-world scenario, since companies

typically invent various products and maintain prod-

uct variants or versions (in re-design processes) over

time. In particular, when re-designing a product, a lot

of documentation on the single parts of the product is

already available and needs to be taken into account.

Especially, at this stage it is very valuable to know

about former design decisions. Therefore, efﬁcient

management and support of the knowledge needs of

people involved in a design process is crucial and get-

ting more and more importance nowadays.

For knowledge-intensive, collaborative processes,

information systems used within the industry today

are insufﬁcient for providing efﬁcient and holistic

support during a process as well as across different

processes. Especially, it is up to individuals to (man-

ually) create and manage their own information space

to increase the efﬁciency of their work process. This

space is neither explicitly represented (and hence not

documented) nor managed by a dedicated system. It

cannot be communicated and not be shared with oth-

ers. Furthermore, support systems are usually pas-

sive, i.e. the user has to repeatedly ask for informa-

tion, rather than getting notiﬁed about relevant infor-

mation whenever it is available. This shifts the bur-

den of information gathering to the end-user rather

than into a “smart” information space and therefore

absorbs a lot of useful attention of the end-user from

the actual problem solving process. This situation is

depicted in Fig. 2.

For this reason, we propose a novel approach to-

Figure 2: The Perspective of an Individual in the Process.

wards developing a generic engineering support envi-

ronment that combines ideas underlying the Seman-

tic Web, Semantic Desktops, Group Memory Sys-

tems and Multi-Agent Systems into a coherent envi-

ronment to overcome the problems and improves the

current situation for individuals involved in engineer-

ing process (see Fig. 2) towards the situation shown in

Fig. 4. Whereas in this paper, we restrict ourselves on

the presentation of the problem context, the identiﬁ-

cation of requirements for a desired support environ-

ment and the derivation of the conceptual ingredients

of our solution, more technical details and a discus-

sion on economic relevance can be found in (Keller

et al., 2007).

2 MAIN CHALLENGES

From our discussion of the problem context in Sec-

tion 1, we derive the following main challenges that

engineering support environments should properly

address: (i) Explicit representation and manage-

ment of information spaces: individuals in the en-

gineering process are knowledge workers. They work

within their personal information space. A prerequi-

site for ﬁnding relevant information inside this space

is the availability of an explicit representation of such

a space. Explicit representation and management by

a dedicated tool is therefore necessary and desirable.

(ii) Organization and access of information spaces

according to suitable mental models (or world

views): Information spaces can get large and ﬁnding

the right information can become a bottleneck. Creat-

ing some structure within the space, allows to deal

with the information more efﬁciently. But suitable

structures depend on the individuals using a space.

Efﬁcient access of information is facilitated by struc-

tures that correspond to an individual’s “natural” un-

WEBIST 2007 - International Conference on Web Information Systems and Technologies

170

derstanding of the domain (i.e. world view or domain

model). Hence, (a) domain models are needed to

structure information spaces, (b) the structure of per-

sonal information spaces should be customizable for

individuals and (c) these models should be machine-

processable to enable machine-support when access-

ing information.

(iii) Support for the maintenance of various infor-

mation spaces & distinction between private and

shared spaces: Information spaces of individuals

are distinct and should not interfere (if not explic-

itly stated). Usually, there is information that is spe-

ciﬁc to an individual and not intended to be shared

with others. Further, there is a lot of information that

is meant to be shared with others for collaborative

tasks. Individuals are therefore interested in includ-

ing such shared information into their personal infor-

mation space whenever it is relevant to them, i.e. on

a demand basis. Hence, (a) information spaces can

be shared or private, and (b) structured by controlled

overlapping for collaboration purposes.

(iv) Pro-active support: information spaces can be-

come very large over time. Therefore, even when be-

ing structured properly, an individual might not be

able to keep track of the (global) state of his infor-

mation space at a speciﬁc moment in time and there-

fore miss relevant information that comes into being

(i.e. from a shared source) or only detect the infor-

mation unnecessarily late. Instead of requiring an in-

dividual to explicitly ask the system for the current

state of the information space at any point in time (to

explore the space), a pro-active environment is desir-

able, that informs individuals about (small) changes

of the current state of their information space. Such

notiﬁcations should be optional (i.e. applied on a on-

demand basis), selective and speciﬁed by the client.

We consider this elementary form of pro-activeness

as a minimal requirement. Beyond it, more complex

forms of pro-activeness might be useful. In general,

this will require detailed task-speciﬁc (and perhaps

company or product speciﬁc knowledge) which is not

reusable across companies. Therefore, the respective

pro-active support functionality will then have to be

encapsulated inside a dedicated software component

(instead of being a generic functionality supported by

the basic system).

(v) Task-centered support: At any moment in time

during the process, individuals are concerned with a

particular, well-deﬁned task. This implies, that only

parts of the overall personal (and global) information

space are relevant. Therefore, it is desirable that all ir-

relevant information is faded out, since (a) individuals

have limited cognitive capacity, (b) they should focus

as intensively as possible on their task (not being dis-

tracted by other things at that moment). Furthermore,

restricting the overall information space to a relevant

subset at a particular moment in time facilitate scala-

bility of algorithms that are concerned with accessing

the domain models (e.g. ontology reasoning). There-

fore, tasks should be used as a central means for scop-

ing within the system, whenever this is possible.

(vii) The system should support the documentation

of design decisions (that have been taken during a de-

sign process throughout time) to generate a corporate

memory: in fact, this aspect has been identiﬁed as one

of the most useful features of support for engineering

tasks in (Gruber and Russell, 1994).

(viii) The system can not replace existing tools, but

rather has to act as a glue between existing ones: peo-

ple are used to work with speciﬁc tools. Often these

tools are highly specialized to support a particular

task (but not the overall process) and have been de-

veloped with a lot of intellectual and monetary ef-

fort. Hence, various forms of legacy data (e.g ex-

isting documents, entries in databases) have to be in-

tegrated into information spaces. These data sources

often have fundamentally different nature (e.g. ﬁle

systems vs. databases), which must be accessed by

different interaction protocols and are distributed over

various the world.

3 RELATED WORK

In the following, we brieﬂy overview some related

work and concepts that have inspired our approach

and that have similar goals.

Semantic Desktops. Semantic Desktops (Sauer-

mann et al., 2005) are a ﬁrst step towards bringing the

Semantic Web on a personal computer: the underly-

ing idea is to use of ontologies, metadata annotations,

and Semantic Web protocols on desktop computers

to enable integration of desktop applications and the

Web, and therefore a much more focused and inte-

grated personal information management as well as

focused information distribution and collaboration on

the Web beyond sending emails. Recently, (Decker

and Frank, 2004) envisioned the concept of a Net-

worked or Social Semantic Desktops as the ultimate

result of a convergence of three very active recent re-

search ﬁelds: Peer-to-Peer Computing, Social Net-

working and Semantic Web. Essentially, Social Se-

mantic Desktops extend the idea of Semantic Desk-

tops by a strong collaborative dimension based on a

highly decentralized infrastructure.

Our proposal canbe seen as a speciﬁc instantiation

and extension of Semantic Desktops, where we (i)

add a task-speciﬁc dimension (e.g. support is strictly

SPINNING A CORPORATE SEMANTIC WEB FOR PRODUCT ENGINEERING

171

based on an explicit representation of a current task

to be performed), (ii) target at a speciﬁc domain and

therefore aim at providing more domain-speciﬁc sup-

port (e.g. an engineering rational framework), and

(iii) use a speciﬁc set of technologies to realize the

system, that do not have to be the standard technolo-

gies used within the Semantic Desktop Community

(e.g. using a Semantic Wiki as a communication /

collaboration channel).

Organizational Memory Systems. Organiza-

tional memory systems (OMS) (Abecker et al., 1998;

Dieng, 2000) have been proposed as a general ap-

proach to enable integration of dispersed and unstruc-

tured organizational knowledge by enhancing its ac-

cess, dissemination and reuse amongst the members

of an organization and the organization’s information

systems. More recently, (Vasconcelos et al., 2000)

proposed to narrow organizational memory systems

to the concept of group memory systems (GMS): sys-

tems to manage heterogeneous and distributed knowl-

edge embedded in business process activities. There-

fore, GMS can be understood as a specialized version

of OMS that deal with knowledge at a smaller scale.

The system proposed in (Vasconcelos et al., 2000) fo-

cuses on capturing and sharing knowledge about in-

ternal competencies, in particular on human “knowl-

edge sources”. They are understood as information

systems in the classical sense, i.e. passive knowl-

edge stores. In contrast, the environment proposed

here (although sharing many objectives with OMS

and GMS) has a strong pro-active nature (besides pro-

viding a passive corporate memory). A dedicated

agent layer allows to add desired domain-speciﬁc sup-

port functionality. Furthermore, we aim at explicit

task-oriented support for individuals in an engineer-

ing process. At the same time, our use case requires

the instantiation of the agent layer for a speciﬁc func-

tionality, namely an engineering rational framework.

In summary, our system in a way tries to combine

the ideas underlying all these approaches into a co-

herent system and to tailor them towards the speciﬁc

desiderata of product engineering. The idea of scop-

ing is central here. Therefore, it embodies a novel

approach towards support for engineering processes

that is based on solid existing work.

4 PROPOSED SOLUTION

So far, we have arguedthat current tools to support en-

gineering processes are not sufﬁcient, identiﬁed main

reasons and extracted the major challenges that need

to be addressed by an environment that aims at pro-

viding holistic support within and across engineering

processes. In this section, we outline our proposal

for an environment, that combines the ideas underly-

ing Semantic Web, Semantic Desktops, Group Mem-

ory Systems, and Agent-based Systems in order to

tackle the challenges discussed in Sec. 2. In this pa-

per, we focus on motivating and identifying the main

elements and conceptual ideas underlying our system

(Sec. 4.1) and discuss their combination to a concep-

tual system architecture (Sec. 4.2).

4.1 Design Principles

A review of the requirements from Sec. 2, leads to the

following conceptual elements of a holistic support

environment: Requirements (i) Explicit representa-

tion and management of information spaces and (ii)

Organization and access of information spaces ac-

cording to suitable mental models (or world views)

((a) and (c)) motivate the use of ontologies (Staab and

Studer, 2004), as formalized, machine-processable

representation of domain models. Ontologies are the

fundamental semantic data model underlying the sys-

tem. Requirement (ii)(b) can be achieved considering

ontology networks (i.e. a collection of ontologies that

are interrelated by suitable mappings, e.g. (Haase and

Motik, 2005)) instead of a single upper-level ontol-

ogy. The formulation of mappings is up to the user

as well as dedicated administrators of the corporate

knowledge space. Clearly, suitable tools should sim-

plify the deﬁnition of mappings as much as possi-

ble (e.g. by means of simple GUI gestures, such as

the Drag-and-Drop metaphor). Clients always oper-

ate on their personal information space through their

own ontology. Already pre-deﬁned (default) ontolo-

gies in an enterprise could serve as a simplifying start-

ing point for the development of a personal ontology.

Requirement (iii) Support for the maintenance of var-

ious information spaces & distinction between private

and shared spaces motivates the development of a

dedicated ontology management component, which is

capable of identifying ontologies, of managing multi-

ple ontologies and of sharing ontologies amongst a

group of people. The latter might involve concur-

rent changes of the ontology. Conceptually, the com-

ponent provides multiple spaces as semantic com-

munication and collaboration channels between all

other components of the system. Fine-grained access-

control and the support of (partial) imports of ontolo-

gies (via mappings) together allow to construct hier-

archies of (partially) overlapping information spaces

and together cover the desiderata (iii)(a) and (b). We

address requirement (iv) Pro-active support by an on-

tology management component that provides publish-

subscribe capabilities on changes of ontologies. The

most relevant changes are the creation or deletion of

WEBIST 2007 - International Conference on Web Information Systems and Technologies

172

instances in an ontology, however, changes at the ter-

minological level should be supported too (e.g. the

change of the description of a particular class). Fur-

thermore, the use of an agent-infrastructure allows to

encapsulated and integrate more speciﬁc and domain-

dependent support functionality into the environment.

An example of such a functionality is an engineering

rational framework. Requirement (v) Task-centered

support is covered by a dedicated domain model: the

explicit representation of possible tasks of individu-

als within an engineering process and the relation of

these tasks to people, to resources and other activi-

ties. For any individual, a speciﬁcation of the cur-

rent task should be available at any point in time.

Subsequently, all components and agents in the en-

vironment will exploit this information. Requirement

(vii) documentation of design decisions is resolved on

the agent layer by means of a suitable set of agents

that collaboratively implement an engineering ratio-

nal framework. Finally, requirement (viii) No re-

placement of existing tools and integration of exist-

ing legacy systems is taken into account by the use

of an agent-based architecture and a speciﬁc type of

adapter components (called Semantic Facades) that

provide a semantic perspective on the data residing

in a legacy data source. Semantic Facades for stan-

dard data sources (such as relation databases, ﬁle sys-

tems or web servers) and data formats (e.g. PDF,

spreadsheet formats, XML, RDF) must be provided

by the environment. All actors outside of the envi-

ronment (e.g. human users, legacy systems or re-

mote support environments) are integrated via dedi-

cated agents. Agents exploit ontologies for commu-

nication. Communication between agents can happen

in two ways: synchronously by means of a standard

message-based paradigm, as well as asynchronously

(and decoupled in time and reference) by means of a

(semantic) space. Instead of hiding knowledge spaces

of individuals inside an agent at the agent layer, we

strive for an explicit dedicated ontology management

component (based on a distributed infrastructure) that

can be accessed outside the agent-framework as well.

4.2 System Architecture

Conceptually, the architecture of the proposed generic

support system consists of three layers (see Fig. 3),

that successively abstract from the actual technical in-

frastructure available to support individuals today:

(i) The Web Layer addresses the problem of how

to retrieve data objects stores in physical data sources.

It implements the Web identiﬁcation scheme for nam-

ing resources and allows to retrieve data elements

given their URL. Technical knowledge about under-

lying (heterogeneous) interaction protocols, distribu-

Figure 3: Conceptual Layering of the Support Environment.

tion of resources and alike are hidden from the upper

layers. (ii) The Semantic Layer essentially provides

a semantic perspective (i.e. a semantic network) on

the collection of resources that are available inside the

environment. Semantic networks are rooted in on-

tologies. This layer provides capabilities to manage

(multiple) information spaces (i.e. ontologies) and

exploits Semantic Facades to integrate semantic net-

works from legacy data sources. The semantic layer

can be seen as a fundamental infrastructure compo-

nent for semantic-enabled collaboration based on a

very simple conceptual abstraction, namely ontolo-

gies. (iii) The Agent Layer provides advanced ser-

vices to the end-user. Here, we speciﬁcally aim at the

implementation of an engineering framework, i.e. a

software system that allows people to document and

efﬁciently retrieve design decisions that have been

taken during engineering processers. Besides the en-

gineering rational framework, many further services

are desirable in practice. They can be integrated in a

modular way at the agent layer. A concrete example

is a system for the management of corporate compe-

tency (Vasconcelos et al., 2000).

5 CONCLUSIONS

To enhance the support for collaborative, concurrent

engineering processes in distributed teams, we pro-

pose a novel approach to create engineering support

systems. Our approach combines concepts underly-

ing the Semantic Web, Semantic Desktops, Group

Memory Systems and Agents into a coherent engi-

neering support environment. A particularly impor-

tant use case for the proposed system is the realization

of an agent-based engineering rationale framework,

i.e. a system that helps engineers to document and

retrieve design decisions for parts of a product. The

proposed environment is able to provide a more ef-

fective and holistic support for distributed, concurrent

engineeringprocesses, with the following features: (i)

SPINNING A CORPORATE SEMANTIC WEB FOR PRODUCT ENGINEERING

173

Figure 4: Using a Semantic-based Environment for Dedi-

cated Human-centered Support.

decreased complexity of ﬁnding and interacting with

the right information, (ii) semantic-based organiza-

tion and access of information spaces, (iii) person-

alized and customizable organization of the informa-

tion space for every individual, (iv) decreased com-

plexity of identifying the right people, (v) pro-active

and (vi) task-centered support for individuals during

the process, (vii) documentation of design decisions

throughout time to generate a corporate memory. The

proposed system changes the situation that individu-

als have to face within the process from the one that

is shown in Fig. 2 to the situation that is illustrated

in Fig. 4. Conceptually, the system transfers the prin-

ciples underlying the Semantic Web architecture into

enterprises. One fundamental principle thereby is to

not replace the existing tools and systems, but rather

to build an agent-based environment which integrates

them gradually and allows stepwise enrichment of

the overall process support beyond the capabilities of

the single tools. Economically, our approach is ex-

pected to provide as measurable advantages (i) a re-

duction of development time and the time-to-market

(especially for re-design processes, which become in-

creasingly important), (ii) reduction of costs, while at

the same time providing (iii) improved process qual-

ity and quality of the resulting design, and (iv) the

construction and management of a cooperate knowl-

edge base. The latter aspect is especially interest-

ing for companies to address the problem of loosing

knowledge of experienced people leaving the com-

pany, since a lot of knowledge is not stored explicitly

in any IT system inside a company. These information

spaces exist mostly in people’s minds today. For new

staff a cooperate knowledge base reduces the effort

required to become familiar with projects and all rel-

evant information for their daily business. We imple-

mented our ideas in a ﬁrst prototype. The prototype

is functionally not complete yet. It focuses on the use

and sharing of ontologies between users. Concern-

ing the legacy data sources, the prototype is capable

of integrating information from various ﬁle systems.

Future work is the extension of the prototype to sup-

port event subscription and notiﬁcation mechanisms

for information spaces, extended support of impor-

tant legacy data sources (e.g. various popular docu-

ment formats, relational databases), as well as the im-

plementation of an agent-based engineering rationale

framework.

ACKNOWLEDGEMENTS

This work has been supported by the Austrian Federal

Ministry for Transport, Innovation, and Technology

within the SEnSE project (FFG 810807).

REFERENCES

Abecker, A., Bernardi, A., Hinkelmann, K., K¨uhn, O., and

Sintek, M. (1998). Toward a technology for organiza-

tional memories. IEEE Intell. Systems, 13(3):40–48.

Decker, S. and Frank, M. R. (2004). The Networked Seman-

tic Desktop. In WWW Workshop on Applic. Design,

Develop. and Implem. Issues in the Semantic Web.

Dieng, R. (2000). Guest editor’s introduction: Knowledge

management and the internet. IEEE Intelligent Sys-

tems, 15(3):14–17.

Franklin, M. J., Halevy, A. Y., and Maier, D. (2005). From

Databases to Dataspaces: a New Abstraction forInfor-

mation Management. SIGMOD Record, 34(4):27–33.

Gruber, T. R. and Russell, D. M. (1994). Generative design

rationale: Beyond the record and replay paradigm.

Haase, P. and Motik, B. (2005). A mapping system for the

integration of OWL-DL ontologies. In Proceed. of the

Intern. ACM Workshop on Interoperability of Hetero-

gen. Information Systems (IHIS’05), pages 9–16.

Keller, U., Reitbauer, A., and Mungenast, R. (2007). Prin-

ciples of SEnSEful Engineering Support Systems.

TechReport at http://www.uwekeller.net/

publications.html DERI TR 2007-01-07.

Sauermann, L., Bernardi, A., and Dengel, A. (2005).

Overview and outlook on the semantic desktop. In

Proceedings of the Workshop on The Semantic Desk-

top at the ISWC 2005 Conference.

Staab, S. and Studer, R., editors (2004). Handbook on

Ontologies. International Handbooks on Information

Systems. Springer.

Vasconcelos, J., Kimble, C., Gouveia, F., and Kudenko,

D. (2000). A Group Memory System for Corporate

Knowledge Management: An Ontological Approach.

In Proceedings of 1st European Conference on Knowl-

edge Management (ECKM), pages 91–99.

WEBIST 2007 - International Conference on Web Information Systems and Technologies

174