PERSONALIZATION IN VIRTUAL ENTERPRISES

Claudio Biancalana, Fabio Gasparetti and Alessandro Micarelli

Department of Computer Science and Automation, Artiﬁcial Intelligence Laboratory, University Roma Tre

Via della Vasca Navale 79, 00146 Rome, Italy

Keywords:

Knowledge management, Retrieval, Elicitation, Information, User modeling.

Abstract:

Each business company collects, produces and exploits for its activities and goals large amounts of infor-

mation. Most of the times this knowledge makes the intellectual capital for creating value and innovation.

Knowledge management (KM) systems aim at manipulating knowledge by storing and redistributing corpo-

rate information that are acquired from the organizations members. In this context, Virtual Enterprises (VE)

plays a crucial role as not permanent alliances of enterprises joined together to share resources and skills in

order to better respond to business opportunities. The representation and retrieval of distributed knowledge

is an important feature that information systems must provide in order to obtain advantages from this kind of

enterprises. PVE (Personalized Virtual Enterprise) is an ongoing research project for developing a system able

to extract and let different business companies access to collective knowledge required to achieve particular

shared goals. In this paper, we report the most important features of this system, especially in the context of

distributed knowledge representation and retrieval.

1 INTRODUCTION

Knowledge management (KM) (Alavi and Leidner,

2001) has been recognized as a fundamental asset

in the global market place. Companies know-how

and the accumulated knowledge must be collected

and made easily accessible to enhance the efﬁciency

and effectiveness of knowledge-intensive work pro-

cesses and competitive advantages. Improve the capi-

talization on existing knowledge assets facilitates the

creation of new knowledge, proﬁt returns and inno-

vation. In spite of the apparent simplicity of the

term, there are not clear deﬁnitions and classiﬁcations

of knowledge management. Some experts described

the purpose of KM essentially as a document man-

agement system. Other experts prefer to focus on

the process of handling unstructured knowledge, or

more in general, technical and organizational initia-

tives to manage structured and unstructured knowl-

edge in order to store and reuse the internal com-

panys knowledge. In this context, Knowledge Man-

agement Systems (KMS) are deﬁned as Information

Technology-based systems to support and enhance

the organizational processes of knowledge creation,

storage/retrieval, transfer, and application (Alavi and

Leidner, 2001). The process to capture and store

knowledge in ad-hoc repositories to be able to quickly

retrieve information according to the user needs and

goals plays a predominant role in KMS. Knowledge

engineers might help to extract knowledge by elicita-

tion activities. Direct elicitation methods such as sto-

rytelling, interviewing and question answering pro-

vide required information directly from domain ex-

perts interviewed by knowledge engineers that know

what knowledge will be elicited. In indirect elici-

tation methods, information is not directly obtained

from domain experts but there is a further step where

knowledge engineers are involved in the analysis of

the results of elicitation sessions. This second ap-

proach is particularly useful when the knowledge en-

gineers have not fully explored the current domain

or some knowledge has been ignored during the di-

rect elicitation sessions. In these circumstances, in-

direct methods help obtaining information that can-

not be easily expressed directly. While elicitation

methods are useful for capturing knowledge from

users or groups working on particular tasks, the large

amount of information stored in paper documents or

databases are paramount sources of rich knowledge

to exploit during companys activities. Examples of

information maintained at the work group level or be-

yond includes: reports, procedures, pictures, video

tapes, technical standards and databases. Many peo-

ple believe that semi-structured or structured infor-

mation play an important role in a companys knowl-

edge management (Gulli and Signorini, 2005). Struc-

tured information is intended to unambiguous and ex-

plicitly represents concepts in formats that describes

each necessary attribute and property, e.g., relational

database table, while unstructured information usu-

581

Biancalana C., Gasparetti F. and Micarelli A.

PERSONALIZATION IN VIRTUAL ENTERPRISES.

DOI: 10.5220/0001842905810584

In Proceedings of the Fifth International Conference on Web Information Systems and Technologies (WEBIST 2009), page

ISBN: 978-989-8111-81-4

ally requires a human interpretation in order to ex-

tract its intended meaning, e.g., natural language doc-

uments, audio, still images, and video. Structured

information allows users to ﬁnd, share and integrate

information with more precision, also with the sup-

port of software agents (Jennings and Wooldridge,

1996). Nevertheless, unstructured information cov-

ers around 80 percent of all corporate information

(Moore, 2002), and even several public available dig-

ital libraries that can be employed for business activ-

ities such as the Web, are mostly composed of un-

structured information. If we consider Virtual Enter-

prises (VE), where strategic alliances amongst non-

competing companies are settled for the accomplish-

ment of speciﬁc goals, there importance of sharing

and recovery of useful structured/unstructured infor-

mation among partners becomes even more crucial A

company may be not completely aware of the knowl-

edge stored by the partners of the VE. Bringing col-

lectively available complementary competencies and

resources is one of the major achievements of a VE.

In this paper we describe PVE (Personalized Virtuale

Enterprise), an under development KMS able to re-

trieve and share knowledge shared by different com-

panies grouped in a VE. One on the most important

features of the proposed system is the ability to man-

age different kinds of information representations and

degree of formality, i.e., structured, semi-structured

or unstructured, providing an uniform access to sev-

eral different information sources. Instead of long

and costly elicitation processes to manually extract

and annotate knowledge, PVE uses machine learn-

ing and information extraction techniques in order to

draw knowledge from the information stored in the

companies internal digital libraries and map it into

the enterprise ontology. User modeling is employed

to adapt the interaction with the information system

in order to personalize the results of humans informa-

tion seeking activities.

2 RELATED WORK

In recent years, knowledge management has become

a focus of attention for many organizations. Knowl-

edge is considered to be the key source for sustain-

able competitive advantage (Nonaka and Takeuchi,

1995), (Davenport and Prusak, 1997). Therefore,

Knowledge management is aimed at locating, cap-

turing, transferring, sharing and creating knowledge

within and across organizations. It will be clear that

conceptual modeling, as developed within the ﬁeld of

KBS construction, provide key techniques for knowl-

edge management (Gaines et. al., 1997), (Schnei-

der, 2000). A characteristic that turns out to be an

advantage over other industries in terms of manag-

ing intellectual capital is that artifacts (documents)

are already captured in electronic form and can eas-

ily be stored and shared. In addition, software en-

gineers often have a friendly attitude towards using

new technology. This means that a software orga-

nization that implements a knowledge management

system could have a good chance to succeed with this

mission. However, this remains a challenging task be-

cause a knowledge management system is more than

just technology. There are only a few published works

about initiatives to manage knowledge in software or-

ganizations, but all of them talk about the difﬁculty

of achieving employees acceptance and implement-

ing the KM system in a way that maximizes the help

provided to its users (Schneider, 2000), (Schneider,

2001), (Brssler, 1999), (Johansson et. al., 1999).

3 KNOWLEDGE INDEXING

Before facing the problem of knowledge retrieval, it

is essential to analyze how the system indexes the

available information, that is, which representation

has been chosen to guarantee an efﬁcient and ef-

fective retrieval phase. In particular, the require-

ments are twofold: it is essential that knowledge is

quickly retrieved by users, and this knowledge accu-

rately satisﬁes the users information needs in terms

of high precision. An indexing system for busi-

ness companies must also be able to deal with dif-

ferent kinds of information representations, from un-

structured documents based on natural language to

ontology-based knowledge and relational databases.

Moreover, it should provide a comprehensive and ho-

mogenous human-computer interface for knowledge

retrieval. In order to provide the aforementioned pre-

requisite, it is necessary to consider different types of

information and the degrees of information richness.

Information based on ontological standards, for in-

stance, expresses relationships between typically non-

structured information, e.g., natural language text,

and meta-data. These meta-data usually state features

or classes related to given peaces of information. A

typical example is the association between a docu-

ment and one particular category in a predeﬁned tax-

onomy. As for information stored in databases, we

have an underlying relational model that clearly states

the semantic meaning of each peace of informative

unit, e.g. price, address, location, etc., and therefore

facilitate the interpretation/recovery process. In or-

der to deﬁne a unique representation that deals with

the different types of available information, i.e., natu-

WEBIST 2009 - 5th International Conference on Web Information Systems and Technologies

582

ral language, ontology-based and databases, we must

deﬁne a subset of shared features that is possible to

generalize, and automatic or semi-automatic methods

and techniques for translating information from one

of these representations to the internal one. Ontol-

ogy languages capture high-quality relationships and

meta-data content that enable logic-based agents to

interpret and take decisions autonomously (Jennings

and Wooldridge, 1996). The Semantic Web extends

this vision to the Internet domain, where information

is no longer based on HTML, but on semantic stan-

dards like OWL (Web Ontology Language), or pre-

liminary standards like DAML+OIL. Nevertheless,

research on these representations is not completed,

and the logical engine able to automatically inter-

preting such information within single or multi-agent

systems with traditional computational resources is

yet to be deﬁned. Moreover, much of the current

available information is written by humans in natu-

ral language so additional effort is required to trans-

late this information in the new ontology-based lan-

guages. Translating large amount of non-structured

information into new formalisms is not an activity that

can be accomplished manually. Research to develop

methods and techniques for this goal has not reached

completely satisfying results. As for unstructured

information based on natural language poses many

problems during the indexing, but it is also particu-

larly problematic to retrieve it efﬁciently. The well-

known vocabulary problem (Furnas et. al., 1987) for

instance, points out further issues in terms of syn-

onymity and polisemy of words that do not allow

users to express univocally their information needs.

3.1 Internal Representation of

Knowledge

As previously stated, the proposed internal repre-

sentation of knowledge deﬁnes some common fea-

tures shared among the three kinds of information

brieﬂy described. This sort of intermediate repre-

sentation consists of traditional non-structured infor-

mation with associated meta-information related to

concepts of a taxonomy of the business domain for

the given virtual enterprise (see ﬁgure 1). In a few

words, each informative unit is classiﬁed in a sub-

set of categories from a simpliﬁed ontology. Such

meta-information can be exploited both in the re-

trieval phase, reducing possible ambiguities in the

processed information, and to re-organize the knowl-

edge in more efﬁcient ways for further user search ac-

tivities, e.g., online hierarchical clustering (Gulli and

Signorini, 2005).

One further advantage of such a representation is

Figure 1: Each informative uint, e.g., a document or a seg-

ment of it, is associated with one or more categories in a

given taxonomy.

the chance to exploit traditional and well-known in-

dexing and retrieval techniques developed in the In-

formation Retrieval ﬁeld, as search engines based on

the Vector space model (Salton and McGill, 1983).

Such systems guarantee quick response time thanks

to data structures appropriately studied for efﬁciently

memorizing the input documents. Additional infor-

mation can be easily indexed and retrieved together

with the original data in appropriate ﬁelds that can be

used during the recovery process. Even though the

proposed representation simpliﬁes the stored meta-

data, i.e., there are no relationships between infor-

mative units such as IS-A or HAS-A relations in

ontology-based languages; the burden is now on pop-

ulating the internal knowledge base given the avail-

able information. In other words we have to deﬁne

techniques and methodologies to transform the infor-

mation represented in one of the three-above men-

tioned typologies, into the proposed intermediate rep-

resentation.

4 EVALUATION TEST BED

We are currently deﬁning the test-bed for the eval-

uation of PVE. We have assumed a 6-month lasting

period, starting from a set-up phase, where we set-

tle a conﬁguration of the enterprise, i.e., number and

typology of the companies, and install the various dis-

tributed modules. Afterwards, the VE prototypes will

be instantiated and evaluated. Due to the nature of

the prototype, i.e., a distributed environment with dif-

ferent kinds of knowledge representations and typolo-

gies of users, we have decided to evaluate the system

in a real scenario. In this way, we are able to eval-

PERSONALIZATION IN VIRTUAL ENTERPRISES

583

uate all the available features through precision and

accuracy measures, and the analysis of ad-hoc user

questionnaires.

5 CONCLUSIONS

The goal of PVE project is studying and devising

methodologies and techniques for distributed man-

agement of knowledge in the context of a Virtual en-

terprise, favoring synergies and the interaction among

different companies. In particular, we have intro-

duced a knowledge indexing and retrieval engine used

for manage the information stored in a company. This

engine is used by the users to retrieve documents re-

lated to the current needs. The retrieval is enhanced

with semantic metadata extracted by means of in-

formation extraction techniques. Moreover, the sys-

tem employs a user modeling component to adapt the

human-computer interaction during information seek-

ing activities.

REFERENCES

M. Alavi and D.E. Leidner, Review: Knowledge Manage-

ment and Knowledge Management Systems: Concep-

tual Foundations and Research Issues. MIS Quarterly,

Vol. 25 No. 1, pp. 107-136, 2001.

F. Baader, D. Calvanese, D.L. McGuinness, D. Nardi and

P.F. Patel-Schneider (Eds.), The Description Logic

Handbook: Theory, Implementation, and Applica-

tions. Cambridge University Press, 2003.

P. Brssler, Knowledge Management at a Software Engi-

neering Company - An Experience Report, Workshop

on Learning Software Organizations, LSO’99, Kaiser-

slautern, Germany, pp. 163-170, 1999.

P. Brusilovsky and M. Maybury, From adaptive hyperme-

dia to the adaptive web. Communications of the ACM,

Vol. 45 No. 5, pp.3033, 2002.

F. Y. Y. Choi, Advances in domain independent linear text

segmentation, In Proc. of the 1st conference on North

American chapter of the Association for Computa-

tional Linguistics. Seattle, Washington, pp2633, 2000.

C. W. Choo, Information Management for the Intelligent

Organization: Roles and Implications for the Informa-

tion professions. In Proc. of the 1995 Digital Libraries

Conference, Singapore, March 1995.

T.H. Davenport and L. Prusak, Working Knowledge: How

Organizations Manage What They Know, Harvard

Business School Press, 1997.

P. Ferragina and A. Gulli, A personalized search engine

based on web-snippet hierarchical clustering. In Proc.

Of the 14th World Wide Web Conference, 2005.

G.W. Furnas, T.K. Landauer, L.M. Gomez, and S.T. Du-

mais, The vocabulary problem in human-system com-

munication. Commun. ACM, Vol. 30 No. 11, pp.

964971, 1987.

B.R. Gaines, M.A. Musen and R. Uthurusamy, Artiﬁcial In-

telligence in Knowledge Management, Technical Re-

port SS-97-01, Stanford University, 1997.

S. Gauch, M. Speretta, A. Chandramouli and A. Micarelli,

User Proﬁles for Personalized Information Access. In

Brusilovsky, P., Kobsa, A., Nejdl, W., eds.: The Adap-

tive Web: Methods and Strategies of Web Personaliza-

tion. Volume 4321 of Lecture Notes in Computer Sci-

ence, pp.54-89. Springer-Verlag, Berlin Heidelberg

New York (2007).

A. Gulli and A. Signorini, The indexable web is more than

11.5 billion pages. In the Proc. of the 14th interna-

tional conference on World Wide Web (WWW05),

pp902903, New York, USA, 2005.

R. Hartley, R. Kelsey and R. Webster, Consolidating Multi-

Source and Multi-Media Knowledge. In AAAI Spring

Symposium Articial Intelligence in Knowledge Man-

agement. AAAI, March 1997.

N. R. Jennings and M. Wooldridge, Software Agents. IEE

Review. Vol. 42 No. 1, pp17-21. January 1996.

T. Joachims, Text categorization with support vector ma-

chines: learning with many relevant features. In Proc.

of ECML-98, 10th European Conference on Machine

Learning, pp.137142, Chemnitz, DE, 1998. Springer

Verlag, Heidelberg, DE. Published in the Lecture

Notes in Computer Science series, number 1398.

1998.

T. Joachims, Transductive inference for text classiﬁcation

using support vector machines. In Proc. of ICML-99,

16th International Conference on Machine Learning,

pp.200209, Bled, SL, 1999. Morgan Kaufmann Pub-

lishers, SF, US, 1999.

C. Johansson, C. and P. C. M. Hall, Talk to Paula and Peter

- They are Experienced, Workshop on Learning Soft-

ware Organizations, LSO’99, pp. 171-185, 1999.

D. Kelly and J. Teevan, Implicit feedback for inferring user

preference: a bibliography. SIGIR Forum, Vol. 37 No.

2, pp.1828, 2003.

G. Lawton, Knowledge Management: Ready for Prime

Time IEEE Computer, Vol. 34, No. 2, pp. 12-14, 2001.

C. Moore, Diving into Data. InfoWorld. October 25, 2002.

I. Nonaka and H. Takeuchi, The Knowledge-Creating Com-

pany, Oxford University Press, New York, 1995.

P. Pirolli, P. Schank, M.A. Hearst and C. Diehl, Scat-

ter/Gather Browsing Communicates the Topic Struc-

ture of a Very large Text Collection, In Proc. of the

ACM SIGCHI Conference on Human Factors in Com-

puting Systems (CHI), May 1996.

G. Salton and M. McGill, An Introduction to modern infor-

mation retrieval. Mc-Graw-Hill, New York, NY, 1983.

K. Schneider, LIDS: A Light-weight Approach to Experi-

ence Elicitation and Reuse, The Profes 2000 Confer-

ence, Oulo, Finland, 2000, pp. 407-424, 2000.

K. Schneider, Experience Magnets - Attracting Expe-

riences, Not Just Storing Them, Product Focused

Software Process Improvement, PrOFES’01, Kaiser-

slautern, Germany, pp. 126-140, 2001.

F. Sebastiani, Machine learning in automated text catego-

rization. ACM Computer Surveys. 34(1): 1-47 (2002).

WEBIST 2009 - 5th International Conference on Web Information Systems and Technologies

584