ELSNET Goes East
(
And Stays There)

Ingrid van LOON and Erik-Jan van der LINDEN

Institute for Language, Logic and Computation, University of Amsterdam,

Plantage Muidergracht 24, 1018 TV Amsterdam, The Netherlands

e-mail: elseast@wins.uva.nl

http://www.fwi.uva.nl/research/illc/ege/elseast.html


Abstract. The European Network in Language and Speech, ELSNET, is an Esprit funded network in the field of Language and Speech technology. It started out as a Western-European network. On the basis of funding from ESPRIT, COPERNICUS, LRE and INTAS, ELSNET has started a process of changing into a Pan-European network. In 1994 ELSNET carried out a survey of organisations in Central and Eastern Europe and the Newly Independent States of the former Soviet Union (CEENIS), that are involved in the Language and Speech technology. On the basis of the findings of this survey, the network formulated a number of projects to experiment with formats for cooperation with CEENIS. One of these projects is the COPERNICUS funded project ELSNET goes East. In this paper, we shortly explain the term Language and Speech Technology, we describe ELSNET and the findings from the ELSNET survey; we discuss the ongoing and future activities of the project ELSNET goes East, and we end with a conclusion1 .

1. LANGUAGE AND SPEECH TECHNOLOGY

The scientific dimension of Language and Speech Technology centres on the way that information is communicated using spoken and written language. Specifically, it is concerned with Natural Language - that is, languages such as French, Hungarian or Japanese which are in everyday use by human beings. Users of language bring sophisticated knowledge to bear on complex cognitive tasks: for example, extracting information from the acoustic signals of Speech, and dynamically updating their knowledge states on the basis of heavily context-dependent utterances.

Our understanding of these processes is far from complete, and progress will require sustained interdisciplinary research, drawing on artificial intelligence and computer science, linguistics, electrical engineering, and cognitive psychology. Nevertheless, rule-based, statistical and connectionist methodologies have all contributed to the development of viable engineering techniques over the last decade. As a result, a wide range of computational applications now incorporate Language Technology.

Coordination on a European scale is absolutely essential for addressing a number of scientific and practical problems. Although European investment and R&D -effort in Natural Language and Speech is comparable to that in the USA and Japan, the fragmentation of effort across Europe threatens to severely diminish the effectiveness of this investment: thus, co-ordinated efforts are essential. However, there are a number of structural impediments to overcome. First, the Natural Language and Speech communities have different methodological frameworks, which obstructs fruitful multi disciplinary cooperation. Second, the language industries lack the basic infrastructure for development and evaluation of technology for the community's languages. Furthermore, the industry suffers from a lack of highly trained professionals: they must be trained in hitherto unrelated disciplines. In addition, many Language and Speech systems rely on ad hoc solutions that work for one application in one language, but that do not generalise well. This is caused by a lack of broadly accepted standards. It is necessary to accelerate progress by increasing the level of cooperation between academic and industrial groups, by helping to reconcile the distinct methodological positions which divide the Natural Language and Speech communities, by encouraging the development of the infrastructure needed to establish standards in order to enable transfer of technologies to the Information Technology industry, and by contributing to a European system for training professionals.

2. ELSNET

ELSNET, the European Network in Language and Speech, was set up in 1991 against this background as a 'Network of Excellence' under ESPRIT Basic Research Actions. ELSNET is funded by ESPRIT until the end of 1999. The primary objective of ELSNET is to support the coordination of European research and training activities in the field of Natural Language and Speech. ELSNET currently consists of about sixty academic and research institutes and over forty industrial affiliates.

During the next few years, ELSNET will address four problem areas:

Much of the business of ELSNET is carried out via so called Task Groups. The following five groups are in place: the Research Task Group, the Training and Mobility Task Group, the Industrial Links Task Group, the Language and Speech Resources Task Group and the Information Dissemination Task group.

3. EXTENSION TO CEENIS: ELSNET GOES EAST

3.1 The 1994 ELSNET survey

In 1994, ELSNET carried out a survey of Natural Language and Speech organisations in CEENIS. About a hundred organisations were included. The most important results of this survey are presented below:

3.2 ELSNET goes East

On basis of the findings of this survey, ELSNET goes East started in 1995 with funding from the COPERNICUS program. In principle, the projects runs for two years until the end of 1996. However, due to under spending, an unfunded extension for the year 1997 is expected. ELSNET goes East is financed as a concerted action. A concerted action 'co-ordinates, across borders, research and development activities which are already underway within the countries'. The idea behind a concerted action is that many technical needs can best be solved by working together in a multi-disciplinary way, and on a multinational level, rather than in a single country or in a particular R&D project.

ELSNET goes East has elaborated this idea by building an infrastructure in CEENIS, which has the following objectives:

(i) create an information infrastructure;

(ii) establish contacts by means of visit;

(iii) improve access to hardware, software and "R&D -ware'';

(iv) allow access to expertise on interdisciplinary research, the Language Engineering market and cooperation between academic and industrial organisations;

(v) improve the training system.

Summarising its objectives: ELSNET goes East aims at an extension of ELSNET to countries in CEENIS, in order to enable ELSNET to play a role in CEENIS similar to the one it plays in Western Europe. The ultimate goal is to set up a Pan-European infrastructure for industrial and academic organisations in the field of Natural Language and Speech.

If the objectives mentioned above are achieved this will have very important benefits, both for the CEENIS itself, the EC, and the pan-European community. Firstly, and most importantly, it will alleviate the isolation of the Language and Speech community in CEENIS countries. Participation in a network with colleagues all over Western Europe and CEENIS will open up a wide spectrum of possibilities. In the longer term it creates a general open pan-European research infrastructure which can serve as the backbone for long-term cooperation between academic and industrial groups, and groups in the area of Language and Speech. Coordination on a pan-European level is of great importance for the area. Research in CEENIS is in a crucial stage of development. Coordination of activities in this stage will contribute to research paradigms and technologies of a true pan-European nature. Without coordination, the fragmentation of effort across Europe mentioned before, could easily be multiplied. Co-ordinated efforts are essential for the pan-European community to compete in this scientific and technological arena.

4. ACTIVITIES IN ELSNET GOES EAST

4.1 Structure

The project has ten coordinating nodes (a node is an organisation), representing the following 12 CEENIS countries: The Baltic States (Lithuania, Latvia and Estonia), Belarus, Bulgaria, Czech Republic, Georgia, Hungary, Poland, Romania, Russia and Ukraine. Contact persons in these nodes serve as a representative for their node and their country, and are responsible for contacts with ELSNET goes East. They have an important role in carrying out the activities mentioned in the next paragraph. In addition, a Steering Committee, consisting of eight members both from the West and the East, takes part in the management of the project and discusses all issues of policy concerning the project.

4.2 Ongoing activities

The objectives mentioned in paragraph 3.2 will be implemented by means of extension of the ELSNET infrastructure, which has shown its use in Western Europe, to CEENIS. The specific actions to accomplish this, are the following:

- ELSNET maintains the electronic mailing list elsnet-list. The total amount of subscribers is about 750, of which approximately 150 subscribers are from CEENIS.

- ELSNET publishes a bi-monthly newsletter, Elsnews. It contains information of a reasonably topical nature. The newsletter is devoted partly to news about ELSNET and its members: new activities, new technical reports and a spotlight on individual nodes, and partly to information about other activities of interest to the ELSNET community, such as recent developments in ESPRIT, LRE, and related EC initiatives. It is sent for free to about 400 addresses, including 130 addresses in CEENIS.

- ELSNET goes East developed its own World Wide Web pages with information about the project, membership to ELSNET, scientific activities in the field, and many other subjects.

4.3 Future activities

The results of the efforts so far are very satisfactory. Recently the project has been reviewed by the European Commission (May 1996), and the evaluation was highly positive. Nevertheless, ELSNET goes East likes to reconsider and extend some of its activities. Plans for the future, based on ideas provided by the the Steering Committee, the co-ordinators and the reviewers, include:

Of course, the main problem in CEENIS is a lack of funding. Unfortunately, a project like ELSNET goes East, can not fund research as such. What we try to do is to build the necessary infrastructure that helps to alleviate this main problem. We hope that the activities of ELSNET and ELSNET goes East will lead to a permanent Pan-European structure that is both profitable for the West and the East. Or as a Russian colleague once stated: 'ELSNET goes East, and stays there'.

CONCLUSION

In this paper, we have indicated some of the problems encountered in the field of Language and Speech in CEENIS, especially where Pan-European cooperation is concerned. We have discussed the ongoing and future activities of ELSNET goes East that try to alleviate these problems. We hope to have shown that, in general, incorporation of research in CEENIS in a network structure can contribute to improving Pan-European cooperation.

BIBLIOGRAPHY

Van der Linden, E., (to appear) Internet for the Language and Speech Community. In: Proceedings of the Networking Awareness Day for the Language and Speech Community. Edited by Alexander Barulin and Vera Semenova.

Van der Linden, E., and Boguslavsky, I., (to appear) The project ELSNET goes East. In: Proceedings of the Networking Awareness Day for the Language and Speech Community. Edited by Alexander Barulin and Vera Semenova.

Boguslavsky, I., and Lazourski A. (eds.) Integration of Language and Speech, Proceedings of the workshop

Barulin, A. and Liberman, M., (to appear) Evaluation of the ELSNET goes East Concerted Action

Appendix PARTICIPANTS IN ELSNET GOES EAST

Institute for Logic, Language and Computation, University of Amsterdam, The Netherlands (Coordination)
Research Institute for Language and Speech, University of Utrecht, The Netherlands
Institute of Formal and Applied Linguistics, Charles University, Czech Republic
Institut für Maschinelle Sprachverarbeitung, University of Stuttgart, Germany
Speech Acoustics Laboratory, Institute of Fundamental Technological Research, Polish Academy of Sciences, Poland
Rank Xerox Research Centre, France
Institute for Information Transmission Problems, Russian Academy of Sciences, Russia
Speech Recognition and Synthesis Laboratory, Institute of Engineering Cybernetics, Belarusian Academy of Sciences, Belarus
Speech Communication Group, Language and Recognition Group, France
Centre for Cognitive Science, University of Edinburgh, United Kingdom
Siemens AG, Germany
Department of Speech Processing, Instituto de Engenharia de Sistemas e Computadores, Portugal
Centre of Cognitieve Science, Roskilde University, Denmark
Speech Communication and Music Acoustics, Royal Institute of Technology, Sweden
Dipartimento di Linguistica, Universit'a di Pisa, Italy
Linguistic Modelling Laboratory, Institute of Informatics, Bulgarian Academy of Sciences, Bulgaria
Research Institute for Linguistics, Hungarian Academy of Sciences, Hungary
Center for Advanced Research in Machine Learning, Natural Language Processing and Conceptual Modelling, Romanian Academy, Romania
Recognition processes Department, Institute of Mathematics and Informatics, Lithuania
Acoustics Research Laboratory, Department of Telematics and Telecommunications, Technical University of Budapest, Hungary
International Laboratory of Intelligent Systems, Russia
MorphoLogic Gmk, Hungary
Glushkov Institute of Cybernetics, National Academy of Sciences of Ukraine, Ukraine
Centre Of Language, Logic And Speech, Tbilisi State University, Georgia


BENEFIT Special Report Main Page