THE SOCIAL ONTOLOGY BUILDING AND EVOLUTION

(SOBE) PLATFORM

Daniela Angelucci

, Alessia Barbagallo

, Tania Di Mascio

and Francesco Taglino

Isitituto di Analisi dei Sistemi ed Informatica “A. Ruberti” (IASI-CNR), Viale Manzoni 30, 00185 Rome, Italy

TXT e-solution, Via Frigia 27, I-20126 Milan, Italy

Keywords: Ontology building, Knowledge extraction, Social participation, Service oriented architecture.

Abstract: In this paper, the Social Ontology Building and Evolution (SOBE) method and the corresponding software

platform for cooperative ontology building in the context of a cluster of enterprises is presented. The SOBE

is characterized by three main aspects: (i) automatic knowledge extraction from unstructured documents; (ii)

social validation, involving a community of domain experts; (iii) a step-wise approach which goes through

five main steps which produce incremental results towards the construction of a domain ontology.

1 INTRODUCTION

Knowledge is one of an organization's most valuable

assets. For this reason, it is crucial to establish

methodologies and technologies to manage the

creation, capture, and sharing of knowledge and

information at and among all levels of the

organization. In particular, developing enterprise

ontologies to support structuring and sharing of

enterprise information is an emerging practice in

enterprise knowledge management.

In this paper, we present the SOBE (Social

Ontology Building and Evolution) platform for

supporting ontology building in the context of an

enterprises cluster. SOBE is strongly rooted on three

main aspects: (i) automatic knowledge extraction

from unstructured enterprise documents which are

pregnant with enterprise knowledge; (ii) social

participation of a community of experts which are

enabled and requested to validate, discuss about (to

find a consensus), and enrich the results of the

automatic extraction; (iii) a step-wise approach that

goes through intermediate results: lexicon, glossary

and taxonomy.

Developing ontologies in general is not a

project undertaken by a single person, but rather it is

a large project with numerous participants; this

approach is known as social participation. A

collaborative approach for solving terminological

problems is presented in (Campbell et al., 1998).

Some years later, a collaborative approach for

ontology building was introduced by (Holsapple and

Joshi, 2002). ased on the Delphi method (Lindstone

and Turoff, 1975), this approach proposes

collaborative development of ontologies in

Agentcities (Ceccaroni and Ribiere, 2002) was

carried out through both face-to-face meetings and

remote communication among several partners of

the EU Agentcities RTD project. Another

contribution on consensus building techniques

applied to ontology engineering is presented by

(Karapiperis and Apostolou, 2006). The proposed

methodology includes the definition of ontology

design criteria, the development of an initial

ontology, the iterative process of ontology

evaluation and evolution, and the ontology

application. Moreover, (Tempich et al., 2007)

proposes the use of argumentation theory, and

(Walton, 2009) proposes the support of ontology

engineering. With respect to (Holsapple and Joshi,

2002), (Karapiperis and Apostolou, 2006), SOBE

deals with the complete ontology evolution process,

the detailed collaborative process (i.e., debating, and

voting) and the needed steps and relative milestones.

Furthermore, in SOBE, we introduce automatic tools

trying to integrate at best human and software-based

activities. With respect to (Campbell et al., 2010)

and (Holsapple and Joshi, 2002), only dealing with

the terminological level, we focus on a substantial

extension and enrichment of an existing ontology

rather than on its simple revision. Finally, in

(Ceccaroni and Ribiere, 2002), the type of

416

Angelucci D., Barbagallo A., Di Mascio T. and Taglino F..

THE SOCIAL ONTOLOGY BUILDING AND EVOLUTION (SOBE) PLATFORM.

DOI: 10.5220/0003101204160419

In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2010), pages 416-419

ISBN: 978-989-8425-29-4

 2010 SCITEPRESS (Science and Technology Publications, Lda.)

communication is inadequate with respect to the

SOBE community dimension.

The objective of this paper is to present the

whole work, by also adding with respect to

(Barbagallo et al., 2010) an overview of the platform

architecture and a case study in the context of the

COIN European project. The rest of the paper is

organized as follows. In Section 2, we describe the

social ontology evolution process on which SOBE is

based. In Section 3, we present the platform

architecture and the modules that perform the

various steps of the process. Section 4 is dedicated

to the presentation of a case of study in which we

carried out an experimentation of the SOBE

platform in the ICT domain. Finally, in Section 5,

we present conclusions and future works.

2 THE SOBE PROCESS

The social ontology building and evolution (SOBE)

process (see Figure 1) exploits the UPON

methodology (De Nicola et al., 2009), and the

ontology learning methodology defined in (Velardi

et al., 2007), enriching them with social participation

aspects. In particular, UPON is characterized by an

incremental nature, reflected by the outcomes

produced in the different phases of the process: first

the relevant terms in the domain are identified and

gathered in a lexicon; then the latter is progressively

enriched with definitions, yielding a glossary;

adding to it the specialisation relationships allows a

taxonomy to be produced, until further enrichments

and a final formalization produces the sought

domain ontology. The SOBE process exploits this

step-wise approach and enriches it through an

automatic support for knowledge extraction from

existing digital resources, and social participation

aspects for consensus reaching among the

community of experts that participates to the

ontology building. The automatic knowledge

extraction support aims at reducing the workload of

the people involved in the ontology building, and at

reusing the amount of knowledge contained in any

type of existing documental resources (e.g.,

technical papers and reports, standards

specifications, etc.), and structured resources (e.g.,

dictionaries, thesauri, ontologies). In accordance

with the UPON methodology, SOBE firstly

addresses the terminological aspects, i.e. the lexicon.

The start up consists in processing a corpus of

documents, related to the addressed domain, for

automatically extracting terms that are considered

relevant in that domain. This extraction phase is

based on natural language processing techniques,

statistical analysis, and contrastive analysis against a

pre-defined corpora of documents related to

different domains. The extracted terms are referred

in Figure 1 as E-Lexicon. In the case of enrichment

of an existing ontology, the E-Lexicon is filtered out

of the terms that are already in the lexicon of the

current ontology (O-Lexicon). Then, the E-Lexicon

is validated by the community of experts to reach an

agreement on the new terms to be included in the

ontology (N-Lexicon). After the identification of the

N-Lexicon, terms have to be enriched with natural

language definitions in order to build the desired

glossary. As in the philosophy of SOBE, definitions

are firstly extracted from existing dictionaries or

ontologies (e.g., Google Define, WordNet), yielding

the E-Glossary which has to be humanly validated.

Glossary validation is performed by voting extracted

definitions. Any potential conflict due to lack of

agreement or terms with no definitions are managed

by opening discussion forums about glossary entries.

The result of the glossary validation step is gathered

in the N-Glossary. The following step is the

categorization of the N-Glossary entries by

associating to each of them a kind (i.e., Object,

Process, Actor) in accordance with the OPAL

framework (D’Antonio et al., 2007). Terms with

definitions and associated kinds represent the new

concepts to be inserted in the ontology. Starting

from the newly acquired concepts definitions,

natural language processing techniques allow an

automatic proposal of their hypernyms, producing a

set of micro-taxonomies: Eµ-Taxonomies. An Eµ-

Taxonomy is a specialization hierarchy between

concepts. In the case of ontology enrichment the Eµ-

Taxonomies are merged with the taxonomy of the

existing ontology (i.e., O-Taxonomy) producing the

N-Taxonomy. In the last step, the taxonomy is

enriched with other relationships (e.g., part of,

attributes) producing the final N-Ontology. The

human validation phase involves three types of

actors, who play specific roles in validating the

results of the automatic extraction tools: the

Ontology Master (OM), Participants (Ps) and

Moderators (Ms). The OM is responsible for the

whole ontology enrichment process and is in charge

of managing and supervising all its different phases.

People involved as

Ps play an active role in the

validation of the extraction tools results. They are in

charge of validating, adding, modifying terms and

definitions, and proposing new ones. This represents

the SOBE social participation aspect which is

realized through three different mechanisms: voting,

discussing and proposing. Voting enables the

THE SOCIAL ONTOLOGY BUILDING AND EVOLUTION (SOBE) PLATFORM

417

classical validation represented by accepting or

discarding terms and definitions proposed by the

automatic knowledge extraction services. Discussing

enables a more active participation of Ps with the

aim of finding a shared agreement on content both in

the lexicon and glossary building steps. The system

provides several mechanisms for managing users’

validations and processing their results. As an

example, after the glossary validation, two scenarios

are possible: (1) all the terms have at least one

definition and all of them have been accepted with a

reasonable consensus (more than 60% of Ps): in this

case the OM is suggested to select accepted

definitions; (2) there are some terms with no

definitions or with definitions with low acceptance

rate: in this case the OM is invited to open a

discussion forum for each of these terms. For each

forum a M is designated by the OM from a list of Ps.

S/he is responsible for managing comments and

conversations and for proposing a final definition for

each term at the end of the discussion phase. The

system supports Ms assignment displaying each P’s

reliability, obtained by calculating the precision of

his/her own validation with respect to the accepted

list of terms.

Figure 1: Overview of the SOBE process.

3 THE SOBE ARCHITECTURE

The SOBE platform is designed as a web application

with a classical three layer architecture where: (i) the

Data Layer contains the different knowledge bases,

that are in accordance with the intermediate results

of the SOBE process; (ii) the Logic Layer contains

the functional modules; (iii) the Presentation Layer

manages the graphical user interface and incoming

requests via web services from external client

applications. In particular, the Logic Layer presents

two main components: the Extraction Subsystem

which provides the automatic support in terms of

Lexicon, Glossary and the

-Taxonomies extraction;

the Validation Subsystem which supports the social

participation of the SOBE users for the Lexicon and

Glossary validation. Furthermore, the Extraction

Subsystem is organized in accordance with an open

service oriented paradigm: it is service oriented,

because the core of the knowledge extraction tasks is

provided by external web services (for instance, in

the current implementation of the SOBE platform,

we integrated the TermExtractor and GlossExtractor

(Velardi et al., 2008) web services); it is open in the

sense that it intends to be flexible and ready to allow

the linkage of additional extraction services in order

to: i) integrate results from different knowledge

extraction services to enrich the automatic support of

the SOBE platform; ii) to exploit different extraction

criteria, capabilities and performances of different

services for using the most suitable ones depending

on the dimension, addressed domain and further

features of the analyzed corpus of documents.

4 CASE STUDY: ONTOLOGY

BUILDING IN ICT DOMAIN

In this section we introduce a preliminary case study

of SOBE in building an ontology about the

competencies and skills of a cluster of enterprises in

the ICT domain. The involved enterprises refer to

the IVSZ (Hungarian Association of Information

Technology Companies, http://english.ivsz.hu)

cluster. The IVSZ cluster gathers almost 300

enterprises, however, in order to test the SOBE

platform we involved 8 enterprises, which represent

a small, but real, subset of the cluster itself. The

involved team was composed by 4 people: 1 OM and

3 Ps. The overall ontology building process took

place approximately in 10 days. After the terms

extraction step, an E-Lexicon of 102 terms resulted.

The OM opened a 3-days social validation phase and

after global terms validation, the N-Lexicon was

composed by 21 terms of which: 10 had an

acceptance rate of 100%, 8 had an acceptance rate of

66% and 3 had an acceptance rate of 33%. In

glossary extraction phase, the average number of

automatically extracted definitions per term was 4,

while the maximum was 8. The deadline for social

validation over the extracted definitions was of 4

days. After the first stage of glossary validation, for

14 terms a definition was identified due to the high

KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development

418

acceptance rate of the Ps. For the remaining 7 terms

a forum was opened. This was due to the fact that in

6 cases no definition had reached a sufficient

acceptance rate while for 1 term no definition was

found through glossary extraction and no one was

proposed during the validation phase. After having

entitled the Ms, the OM set a deadline of 3 days for

allowing discussions in forums. After the

categorization of the couples term-definition, we

obtained 3 concepts identified as Actor, 9 as

Process, and 9 as Object. Referring to the taxonomy

phase, in 18 cases the system was able to

automatically extract at least one hypernym. Since

the number of automatically extracted hypernyms

accepted by the OM was 15, in 6 cases hypernyms

had to be manually defined. At the end of the use

case we interviewed the system users to receive

some feedback. They found the system very easy to

use thanks to its user friendly interface and its use

articulated in incremental steps. The social aspects

of the platform granted a distribution of

responsibilities among all the system users and a

consequent drop of the stress level in performing the

tasks. This led to a more participative and

propositive attitude of the users in building the

ontology.

5 CONCLUSIONS

In this paper, the Social Ontology Building and

Evolution process and platform have been presented.

The SOBE platform has been experimented in

the construction of an ontology on ICT

competencies for a cluster of enterprises. The results

have been reported step by step. Our aim for future

works, is to evaluate the ontology generated through

the SOBE process by using it for creating semantic

profiles for the cluster enterprises, i.e. tagging each

enterprise document(s) with concepts from the

ontology, and asking people belonging to the

reference domain to judge the ontology accuracy in

representing the real enterprise competencies.

Moreover, we intend to reinforce the µ-Tax

extraction module for the identification of micro

taxonomies and suggestions on how to attach such

micro taxonomies to the specialization hierarchy of

the existing ontology in the case of ontology

evolution. Furthermore, we intend to address the

automatic identification of synonyms among

extracted terms, with the aim of having, for each

concept in the ontology, one preferred term and a list

of terminological expressions referring to the same

concept.

ACKNOWLEDGEMENTS

This work has been partly funded by the European

Commission through ICT Project COIN:

Collaboration and Interoperability for networked

enterprises (No. ICT-2008-216256). The authors

wish to acknowledge the Commission for its

support.

REFERENCES

Barbagallo A., De Nicola A., Missikoff M., eGovernment

Ontologies: Social Participation in Building and

Evolution. In the proceedings of the 43rd Hawaii

International Conference on System Sciences

(HICSS2010). Koloa, Kauai, Hawaii. January 05-08

2010. ISBN: 978-0-7695-3869-3

Campbell K., Oliver D., and Shortliffe D., The Unified

Medical Language System: toward a collaborative

approach for solving terminologic problems. www.

ncbi.nlm.nih.gov/pubmed/9452982. 1998

Ceccaroni L. and Ribiere M.; Experiences in Modeling

Agentcities Utility-Ontologies with a Collaborative

Approach. In Proceedings of the workshop AAMAS

2002 – W10: Ontologies in Agent Systems (OAS2002),

Bologna, Italy, 2002.

D’Antonio F., Missikoff M., Taglino F., Formalizing the

OPAL eBusiness ontology design patterns with OWL,

In Proc. of I-ESA 2007 conference.

De Nicola, A., Navigli, R., Missikoff, M., A software

engineering approach to ontology building,

Information Systems, 34 (2), 258-275, 2009.

Holsapple, C. W., Joshi, K. D. A collaborative approach to

ontology design. In Communications of the ACM, 45

(2), 42-47, 2002.

Karapiperis S. and Apostolou D. Consensus building in

collaborative ontology engineering processes, Journal

of Universal Knowledge Management 1 (3), 2006.

Lindstone, H. and Turoff M., The Delphi Method:

Technology and Applications. Addison-Wesley,

Reading, MA, 1975.

Noy F. N., Chugh A., Liu W., Musen M. A., A

Framework for Ontology Evolution in Collaborative

Environments. LNCS 4273, pp. 544-558, 2006.

Velardi P., Cucchiarelli A., Petit M.; A taxonomy learning

method and its application to characterize a scientific

web community IEEE Transactions on Knowledge

and Data Engineering, 2007

Velardi P., Navigli R., D'Amadio P.; Mining the Web to

Create Specialized Glossaries. IEEE Intelligent

Systems, 2008

Walton D. Argumentation Theory: A Very Short

Introduction. In “Argumentation in Artificial

Intelligence”, SPRINGER, 2009.

THE SOCIAL ONTOLOGY BUILDING AND EVOLUTION (SOBE) PLATFORM

419