Dagstuhl Seminar: Machine Learning for the Semantic Web

Dagstuhl Seminar

Machine Learning
for the
Semantic Web
(05071)

13-18 February 2005

This bird is a Weka, which is also the name of a well-known suite of machine learning tools.

Dagstuhl-MLSW is sponsored by:
Advanced Knowledge Technologies		Science Foundation Ireland

[ Who's Who? · Schedule · Proceedings · Invitations · Call for Abstracts · Seminar Description · Organizing committee ]

Invitations

The seminar invitation process and registration are now closed. As Dagstuhl has a limited physical capacity, the seminar is now fully booked, and it is no longer possible to register, even if you recieved an invitation. Please note that, as with all Dagstuhl Seminars, participation is strictly by invitation only.

Call for Abstracts

All seminar participants are requested to submit a 2-4 page extended abstract or position paper describing their work. Ideally, the abstract will explicitly highlight the relevance of the research to this intersection of Machine Learning and the Semantic Web. These abstracts will not be reviewed, and there is no time for a draft / camera-ready iteration. Therefore, please submit a polished abstract that is ready for distribution.

There are no fixed style guidelines, but please format your papers for A4 paper. (In Latex2e, "\documentclass[a4paper]{article}" should work.) Also, please submit your paper as PDF. (To convert other formats to PDF, try createpdf.adobe.com.) Submit your PDF file by email as follows:

        Subject: Dagstuhl-MLSW abstract

(Please use this Subject header to ensure that your message doesn't get eaten by Nick's spam filter.)

These abstracts will be assembled into an informal seminar proceedings to be distributed at the event, but we do not plan to formally publish them. So, feel free to recycle text from previously-published articles, or submit your abstract to other conferences or journals for formal publication.

~~DEADLINE: Abstracts are due by FRIDAY 14 JANUARY 2005.~~ -- too late!

Seminar Description

OVERVIEW: The Semantic Web has attracted great attention since the vision was first articulated several years ago. In a nutshell, the Semantic Web will augment conventional Web content with explicit machine-processable semantic metadata, enabling a variety of automated content manipulation and aggregation.

As demonstrated by the Second International Semantic Web Conference (11/2003), the initial "futuristic vision" has matured into a carefully crafted set of substantive technical proposals, such as the Resource Description Framework (RDF) and the Web Ontology Language (OWL).

However, it is widely recognized the Semantic Web will never "take off" until a critical mass of semantic metadata has been deployed. Many SW researchers have therefore built various tools to help developers attach semantic metadata to their content.

More ambitously, machine learning and other artificial intelligence techniques are being developed that generate the requisite semantic metadata in a semi-automated or even entirely automated fashion. For example, machine learning algorithms for information extraction allow large legacy text repositories to be rapidily enriched with semantic metadata, and machine learning approaches to ontology learning and matching are being developed for the Semantic Web context.

The goal of this seminar is to assemble the leading researchers who work at the intersection of machine learning and the Semantic Web, in order to review progress and identify the most significant opportunities and challenges over the next several years. We will also invite leading figures from the "conventional" (hand-crafted metadata) Semantic Web community, to ensure both that our technology is fully appreciated by the Semantic Web community, and that the machine learning community focuses on important and realistic problems.

SPECIFIC TOPICS: The seminar will focus specifically on the following five topics:

Automated document annotation;
Ontology learning and maintenance;
Ontology mapping and merging;
Service discovery; and
Content cleaning and normalization.

These areas represent the major opportunties for machine learning to impact the Semantic Web. The organizing committee has been carefully organized to reflect each of these areas.

STRUCTURE: We will have two days of standard presentations (about 20 short talks by selected participants, as well as two longer invited talks). These two days will be followed by two and a half days of activities not associated with traditional workshops/conferences.

First, we will spend one and a half days working in small groups to discuss the feasability of integrating complementary research prototypes and chart a course for future collaboration. For example, a data-cleaning component could remove the noise that innevitably arises in the output of an automated information extraction system. In some cases, the result may be standard interfaces between components. In other cases, the act of trying to integrate components will expose new research challenges. It may even be possible to actually glue together components into novel demonstrator prototypes.
Second, we will spend one day collectively writing a "white paper" that lays out what we see as the key opportunities/challenges for machine learning and the Semantic Web over the next several years; this White Paper will be published in a leading journal in order to reach a wide audinece and impact the state of the art.

RELATION TO PREVIOUS DAGSTUHL SEMINARS: There have been a number of related seminars. 00121 was one of the first events to explore the ideas of the "Semantic Web", long before that phrase came to be widely used. 02061 dealt with markup languages for the Semantic Web. 02181 and 99261 dealt with information/data integration, which is a key application area for the Semantic Web. 04391 will deal with interoperability and integration in Semantic Web and Web Service contexts. While obviously related, none of these seminars has specifically addressed the issue of how machine learning and other artificial intelligence techniques can be used to bootstrap the Semantic Web by leveraging legacy content and services.

Organizing committee (email addresses mutilated to fool spam-bots)

Fabio Ciravegna · University of Sheffield · t: +44-114-22-21800 · f: +44-114-22-21810 · e: f.ciravegna at dcs . shef . ac . uk · w: www.dcs.shef.ac.uk/~fabio.

AnHai Doan · University of Illinois at Urbana-Champaign · t: +1-217-644-6381 · f: +1-265-6494 · e: anhai at cs . uiuc . edu · w: anhai.cs.uiuc.edu.

Craig Knoblock · Information Sciences Institute / University of Southern California · t: +1-310-448-8786 · f: +1-310-822-0751 · e: knoblock at isi . edy · w: www.isi/edu/~knoblock.

Nicholas Kushmerick (Primary contact) · University College Dublin · t: +353-1-706-2479 · f: +353-1-269-7262 · e: nick at ucd . ie · w: kushmerick.org/nick.

Steffen Staab · University of Karlsruhe · t: +49-721-608-4751 · f: +49 721 608-6580 · e: staab at aifb . uni-karlsruhe . de · w: www.aifb.uni-karlsruhe.de/~sst.