EACL NEWSLETTER

Issue 5

May 2004

Table of Contents
  1. Editorial
  2. EACL Foundation
  3. PASCAL Network of Excellence Challenges
  4. National CL/NLP Conferences: SEPLN Spain
  5. Conference Report: IJCNLP Hainan
  6. Workshop Report: EAMT-04 Malta
  7. EACL Sponsorship Policy
  8. ACL 2004: List of Accepted Papers
  9. Conference Calender

Editorial

An interesting conference season has started for the EACL part of the world. The ACL Barcelona decisions have been made (papers accepted for the main session are listed below) and the COLING Geneva decisions are expected any moment now. In addition, there is LREC in Lisbon, and various smaller events (listed in the calendar below). Some events have already taken place: the EAMT workshop in Malta (report below), and the IJCNLP meeting at Hainan (also reported on below).

In addition, we have an article describing the PASCAL Network challenges, from Ido Dagan. In the series of newsletter items on national CL/NLP events we have Spain (SEPLN). EACL secretary John Carroll announces the establishment of the EACL Foundation, and we also include the newly drawn up EACL policy on sponsorships; information about both of these is online at the EACL web site http://www.eacl.org/

As always, contributions for the next issue of the newsletter are welcomed at vannoord@let.rug.nl.

Gertjan van Noord
Editor

EACL Foundation

The EACL Foundation was established on 29 October 2003, with the official seat in Groningen, The Netherlands. The official title of the foundation is "Stichting European Chapter of the Association for Computational Linguistics", or "Stichting EACL" for short. The motivation for the foundation is to create a "European legal person" in order to allow EACL to apply for grants to European funding bodies.

Important points in the foundation statutes are:

The foundation statutes are online at the EACL web-site http://www.eacl.org/. The EACL Board is very grateful to John Nerbonne for the work he has done towards creating the foundation.

John Carroll
Secretary of the EACL

PASCAL Network of Excellence Challenges

The PASCAL FP6 network of excellence (http://www.pascal-network.org) started its activities officially on December 1, 2003. PASCAL (Pattern Analysis, Statistical Modeling and Computational Learning) involves more than 250 researchers and PhD students from 58 institutions in Europe, Israel and Australia. Its primary objective consists in improving the interactions between experts in statistics, computational learning theory, optimization and a number of other areas to which such theoretical disciplines can be applied, including NLP (program manager Nicola Cancedda) and Textual Information Access (TIA) (program manager Dunja Mladenic).

In this context, a workshop titled "Learning Methods for Text Understanding and Mining" was held on January 26-29, 2004 at the Xerox Research Centre Europe in Grenoble, France. About 70 participants attended tutorials, scientific presentations and the presentation of "challenge" benchmark proposals (see http://www.pascal-network.org/ Workshops/LMTUM04/, including online links to all papers under the Programme section).

PASCAL Challenges are common benchmarks to monitor progress in NLP and TIA, organized by PASCAL members. Detailed announcements of these challenges will come out in the next couple of months and will be available on the PASCAL web site. Three out of the eight Pascal Challenge proposals that were selected for 2004 are in the NLP area. The results of the first two challenges below will be gathered at the PASCAL Challenge workshop in Southampton in the beginning of November 2004 while the last challenge will be launched at that moment. For further information, please contact directly the authors or look at the challenge page of the PASCAL web-site.

Assessing ML methodologies to extract implicit relations from documents
Contact author : fabio@dcs.shf.ac.uk
Start: June 2004, End: October 2004

The challenge is to assess the current situation of Machine Learning algorithms for Information Extraction from documents, identifying future challenges and foster additional research in the field. Aims are to (1) define a methodology for fair comparison of ML algorithms for IE; (2) define a publicly available resource for evaluation that will exist and be used beyond the lifetime of the challenge; (3) perform actual tests of different algorithms in controlled situations so to understand what works and what does not.

Recognizing Textual entailment
Contact author : dagan@cs.biu.ac.il
Start: June 2004, End: October 2004

The goal of this challenge is to create a new benchmark task dedicated to textual entailment - the task of recognizing whether the meaning of one text fragment is entailed by another text or not. This task generalizes prominent semantic inferences which are common within many practical NLP applications, such as Question Answering, Information Extraction, summarization and multi-document summarization, translation evaluation, paraphrasing, and certain types of queries in Information Retrieval. The goal of the challenge is to introduce a generic practical semantic inference task, which will promote the creation of generic semantic models, learning algorithms, knowledge bases and engines that can be utilized for multiple applications.

Ontology evaluation and population
Contact author : eric.gaussier@xrce.xerox.com
Start : November 2004, End : March 2005

Ontologies are formal, explicit specifications of shared conceptualizations, representing concepts and their relations that are relevant to a given domain of discourse. Currently, ontologies are mostly developed as well as used through 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. As human language is a primary mode of knowledge transfer, linguistic analysis of relevant documents for this purpose seems a viable option. More precisely, automation of ontology construction (ontology learning) and use (ontology population through knowledge markup) can be implemented by a combined use of linguistic analysis and machine learning approaches for text mining. This challenge aims at (a) evaluating in a quantitative manner how useful or accurate the extracted ontology classes, attributes and instances are, and (b) developing a well-grounded learning framework for the task. The first task of this challenge (November 2004-March 2005) will deal with Taxonomy Induction and Population while the second task (June 2005-November 2005) will deal with extraction of non taxonomic relations.

Ido Dagan

National CL/NLP Conferences: SEPLN Spain

The Spanish Society for Natural Language Processing (SEPLN) is a not for-profit association, composed of full-member partners and Institutions, which was created in 1984 aiming to promote and disseminate all type of activities related to teaching, research and development in the field of Natural Language Processing, both at the national and international levels. Currently, the Society has around 350 members.

SEPLN is focused on establishing information and scientific materials exchange channels, organising seminars, symposiums and conferences, promoting publications and collaborating with other related Institutions. Members of the society are actively involved in ACL, and have supported the organization of both the ACL/EACL 97 Conference in Madrid, and this year ACL Conference in Barcelona.

Some of the most relevant activities conducted by SEPLN include the celebration of an annual Conference as a meeting point for the range of research groups that are operational in the Natural Language Processing (NLP) field mainly in Spain and Ibero-American countries; the edition of a publication which, counting with an editorial board, assures stable criteria in terms of quality and periodicity; an Internet server offering NLP related information and an e-mail service which, in addition to offering NLP news, operates as a discussion forum for members.

The SEPLN publishes two issues per year of the journal "Procesamiento del Lenguaje Natural". One including papers of the Conference and the other including presentations of current projects, papers, reports of bibliography and summaries of doctoral thesis. (31 issues have been already produced, cf. http://www.sepln.org/revistaSEPLN/pubrevista.htm). An annual award for young researchers was established in 2002, the selected works are published in a special monographic collection, currently three titles are available.

This year the XX Conference of the Spanish Society for Natural Language Processing (SEPLN'04) is held in Barcelona, July 21. Papers can be submitted in all the official languages of Spain as well as English. You can obtain further information about SEPLN at http://www.sepln.org.

M. Felisa Verdejo

Conference Report: IJCNLP Hainan

In March, the Asian counterpart of the EACL organised its first international conference on the tropical Chinese island Hainan. The conference was organised at a holiday resort situated at the beach in a beautiful part of the island close to Sanya City. This exotic location is perhaps an additional reason for the success of the conference: with 226 registered conference attendees the conference did better than expected. The main conference program attracted 211 submissions, which resulted eventually in 95 contributions (about a third in the form of posters). It was a truly international conference with participants from 19 different countries. Most participants were from Asian countries (80%), but there were participants from Europe and USA as well (each about 10%).

The conference was very well organised by the Chinese Information Processing Society of China, with important help from the Chinese University of Hong Kong. It's hard to judge the quality of the conference, but I definitely heard a number of very interesting papers (and yes, there were also a few less interesting papers, which provided good excuse to take a dive in the South Chinese Sea; my ignorance with respect to Asian languages was mostly to blame for those though). Personally, I liked the accompanying workshop entitled 'Beyond Shallow Analyses', which was the reason for my presence in Hainan. It was a small workshop, but very interesting. The invited talk by Mark Johnson during the main conference fitted nicely with the topic of the workshop. Another interesting event was a forum on the preparation of various (very ambitious!) Chinese NLP projects which are being organised in the context of the Beijing 2008 Olympic Games.

At the end of the conference there were two surprises: the best paper award was *not* awarded, because the committee could not choose between the three nominated candidates! The second surprise was the announcement that the next IJCNLP conference will most probably be held at the Korean island Jeju (in october 2005), and be organised by Jong Hyeok Lee of the Pohang University of Science and Technology.

Gertjan van Noord

Workshop Report: EAMT-04 Malta

The fourth workshop of the European Association for Machine Translation was held in Malta on 26th and 27th April 2004. About 55 people attended the workshop, and during the densely packed two days, some eighteen papers were presented at the Foundation for International Studies, the original seat of the University of Malta situated in the centre of Valletta.

About half the papers were oriented towards the special workshop themes: MT-related issues concerning semitic languages, and other languages of EU accession states. Although the subject matter was rather varied, a persistent difficulty with these languages that came across in the talks is lack of appropriate language resources for building MT systems. Some papers addressed these issues directly, and it is interesting to see different approaches being proposed (e.g. the use of machine-learning techniques to infer transfer rules from limited data, the investigation of very closely related accession languages such as Czech and Slovak in order to capitalise on very simple automated techniques).

The invited speaker was Joris Goetschalckx, who, apart from having a last name that is challenging to pronounce, is head of the European Commission Translation Field Office in Malta. He gave some insight into the peculiar situation that obtains in Malta where the decision to make Maltese an official EU language took everyone by surprise, especially the many bilingual speakers who discovered that their knowledge was not up to EU standards. Consequently, where previously there was no problem, there are now two: an official problem of translating large volumes of EU†legislation into Maltese, and an official lack of qualified translators. Both are being addressed, and both are, in different ways, likely to have a positive impact upon MT research carried out locally.

Those wishing to access further details of the workshop can obtain a copy of the proceedings by contacting Mike Rosner, the local organiser (mike.rosner@um.edu.mt) or visiting the EAMT website (www.eamt.org).

Mike Rosner

EACL Sponsorship Policy

Scale

Each year, the EACL may make available a limited number of sponsorships each worth a relatively small amount of money. Sponsorships are awarded only if the EACL budget allows it.

As an indication, in 2003, the EACL awarded three sponsorships, of approximately 750 Euros each.

Focus

In awarding sponsorships, EACL focuses on education in computational linguistics in the geographical area of the EACL. Priority is given to students from Eastern Europe and more generally, to students from countries with hard currency problems (within the geographical area covered by EACL).

In the past, EACL has sponsored introductory courses at European summer-schools (in the form of a contribution towards the presenters' expenses), participation at summer-schools (funding tuition fees and subsistence expenses for students who would otherwise not have been able to come) and participation at student workshops at EACL conferences (contribution towards student-presenters' expenses).

Visibility

In return for sponsorship, EACL expects some visibility (for instance, the sponsorship is announced in a workshop programme and website etc.). A request for sponsorship must include a description of the visibility for EACL generated by the sponsorship.

Procedure

A request for sponsorship should identify a concrete purpose. Indeed, EACL will not sponsor a school or workshop in general, but it will sponsor a particular course, tutorial, etc., or it will sponsor participation for a particular group of students etc.

A request for sponsorship should contain the following information:

When the sponsorship is requested in view of funding the participation of individuals (students, scholars, lecturers etc.) to the event, the selection of the sponsorship recipients will be done jointly by the EACL and the requesting party as follows:

Requests for sponsorships should be directed by email to one of the EACL board members, who passes on the request to the board.

The EACL board decides within a month on a sponsorship request. Sponsorships are normally requested by and awarded to conferences, workshops, summer schools etc. Sponsorship requests from individuals will not be considered.

ACL 2004: List of Accepted Papers ---------------------------------