ECAI-2004 Workshop on Ontology Learning and Population

Towards Evaluation of Text-based Methods in the Semantic Web and Knowledge Discovery Life Cycle

A Workshop at the 16th European Conference on Artificial Intelligence
August 22nd / 23rd 2004
Valencia, Spain

 

 

Download Proceedings

 

Download Presentations

 

 

 

 

Program

Endorsed By

 

Sunday, August 22nd

 

15.30

-

15.45

Welcome / Introduction

15.45

-

16.15

Quantitative and Qualitative Evaluation of the OntoLearn Ontology Learning SystemPresentation

Roberto Navigli, Paola Velardi, Alessandro Cucchiarelli, Francesca Neri: Universita La Sapienza Roma & Universita Politecnica delle Marche

16.15

-

16.45

A Task-based Approach for Ontology Evaluation 

Robert Porzel, Rainer Malaka - European Media Laboratory, Germany – Presentation

16.45

-

17.30

Invited Talk: Ontology Learning and Populating in Genomics

Claire Nedellec: INRA, France

 

Two main Machine Learning approaches are applied to ontology building  from corpus:

-         conceptual clustering for learning the generality relation (is-a) from cooccurrences. No annotation is needed beforehand. However, the scope of the learning target is difficult to control. The validation of the result is done by hand and time-consuming.

-         learning patterns from labeled examples for extracting instances of given ontological relations. The annotation of the learning examples is time consuming. However, the quality and the scope of the learning results are predictable and the validation is done by computer.

We will illustrate the complementarities, the advantages and limitations of these two approaches on the popular problem of ontology learning in functional genomics on a subtle bacterium, Bacillus subtilis.

17.30

-

18.00

Tea Break

18.00

-

18.30

Measuring the Specificity of Terms for Automatic Hierarchy ConstructionPresentation

Pum-Mo Ryu, Key-Sun Choi: KAIST, Korea

18.30

-

19.00

Discovering Knowledge in Texts for the Learning of DOGMA-Inspired Ontologies Presentation

Marie-Laure Reinberger, Peter Spyns: University of Antwerpen & Vrije Universiteit Brussel

19.00

-

19.30

Learning Taxonomic Relations from Heterogeneous Evidence Presentation

Philipp Cimiano, Aleksander Pivk, Lars Schmidt-Thieme, Steffen Staab: University of Karlsruhe & Jozef Stefan Institute, Ljubljana & University of Freiburg

 

 

Monday, August 23rd

 

9.00

-

9.15

Welcome / Introduction

9.15

-

9.45

Precision and Recall for Ontology Enrichment

Andreas Faatz, Ralf Steinmetz: Darmstadt University of Technology

9.45

-

10.15

Extracting Ontologies from Software Documentation: a Semi-Automatic Method and its EvaluationPresentation

Marta Sabou: University of Amsterdam

10.15

-

10.45

Ontology Express: Non-Monotonic Learning of Domain Ontologies from TextPresentation

Norihiro Ogata, Nigel Collier: Osaka University

10.45

-

11.15

Coffee Break with Poster Presentation

Motivation for "Ontology" in Parallel-Text Information Extraction

Mary McGee Wood, Shenghui Wang: University of Manchester

11.15

-

11.30

Introduction Working Session

11.30

-

13.00

Working Session

Towards Evaluation of Ontology Learning and Population – Setting up Guidelines, Metrics and Procedures

13.00

-

13.30

Conclusions / Wrap-Up

 

 

Topic and Motivation

 

 

Ontologies are formal, explicit specifications of shared conceptualizations, representing concepts and their relations that are relevant for a given domain of discourse. Currently, ontologies are mostly developed (including ontology construction, extension, mapping and merging) as well as used (ontology population through knowledge markup) by a manual process, which is very ineffective and may cause major barriers to their large-scale use in such areas as Knowledge Discovery and Semantic Web. The expected central role of ontologies in the organization and functioning of the Semantic Web has been well documented in recent years. Somewhat less traditional is the role of ontologies in incremental approaches to Knowledge Discovery, in which ontologies and machine learning methods are used in combination to mine, interpret and (re-)organize knowledge.

 

As human language is a primary mode of knowledge transfer, linguistic analysis of relevant documents for ontology learning and population seems a viable option. More precisely, automation of these tasks can be implemented by a combined use of linguistic analysis and machine learning approaches for text mining. The workshop will therefore be concerned with reports on the development of such methods, but specifically also with the quantitative evaluation of these methods.

 

Automatic methods for text-based ontology learning and population have developed over recent years (e.g. results from the ECAI-2000, IJCAI-2001, ECAI-2002 workshops on Ontology Learning and the KCAP-2001, ECAI-2002, KCAP-2003 workshops on Knowledge Markup / Ontology Population), but a remaining challenge is to evaluate in a quantitative manner how useful or accurate the extracted ontology classes, properties and instances are. In fact, this is a central issue as it is currently very hard to compare methods and approaches, due to the lack of a shared understanding of the task at hand. The core theme of the workshop therefore will be to develop such a shared understanding through the definition of a clear task (and corresponding sub-tasks), identify resources needed for the task/sub-tasks and to discuss how best to develop an open source evaluation platform.

 

 

Areas of Interest

 

Submissions are invited on these topics in Ontology Learning and Population (OLP):

 

*  Evaluation Methodologies and Metrics for OLP

-

Including Experience and Best Practice from Related Evaluation Efforts in the Context of CLEF, TREC, SENSEVAL, etc.

*  Datasets and Resources for the Evaluation of OLP

*  Definition of Sub-Tasks for OLP

-

Extraction of Taxonomy, Class-hierarchy

-

Extraction of Class-properties, Relations

-

Extraction of Class-instances, Individuals

*  Definition of Related Tasks

-

Ontology Extension, Evolution

-

Ontology Mapping

-

Ontology Merging

*  Text-based Approaches for OLP, for instance (Combinations of):

-

NLP and Linguistic Analysis for OLP

-

(NLP-based) Text-mining for OLP

-

(Ontology-aware) Information Extraction for OLP

*  OLP in the Context of the Semantic Web

*  OLP in the Context of Knowledge Discovery

 

Organizing Committee

 

Paul Buitelaar  (DFKI)         

paulb@dfki.de

Siegfried Handschuh (AIFB)

sha@aifb.uni-karlsruhe.de

Bernardo Magnini (IRST)

magnini@itc.it

 

Program Committee

 

AIFB

Siegfried Handschuh, Steffen Staab, York Sure

Bar Ilan University

Ido Dagan

DFKI

Paul Buitelaar, Andreas Eisele, Michael Sintek

IRIT, Toulouse

Nathalie Aussenac-Gilles

IRST

Bernardo Magnini

Josef Stefan Inst.

Marko Grobelnik

KDLabs

Jörg-Uwe Kietz

LOA-CNR

Aldo Gangemi

MIG-INRA

Claire Nedellec

NCSR “Demokritos”

Georgios Paliouras

Partners HealthCare

Vipul Kashyap

Univ. Antwerpen

Walter Daelemans

Univ. Basque Country

Eneko Agirre

Univ. Paris 13, LIPN

Adeline Nazarenko

Univ. Poly. Madrid     

Asuncion Gomez-Perez

Univ. Roma “La Sapienza

Paola Velardi

Univ. Roma “Tor Vergata

Roberto Basili

Univ. Saarland

Thierry Declerck

Univ. Sheffield

Fabio Ciravegna, Hamish Cunningham, Yorick Wilks

USC/ISI

Eduard Hovy

XRCE

Eric Gaussier

 

Workshop Attendance and Registration

 

All workshop participants must register for ECAI-2004