NEWSLETTER 
No. 1

Contents

• Editorial
• What does TELRI mean?
• TELRI Partners
• TELRI Working Groups
• TELRI Events

EDITORIAL

Language engineering is the core of information technology, and information technology will be the key industry of the next decades. The information super highways conceived today will transport a variety of data, images, sounds, tables, figures, calculations, and process protocols. To make these data intelligible, they must be bound together by language. Without natural language processing information remains incomprehensible. More than any other continent, Europe is multilingual. This situation provides a challenge to European language technology. We all want information to cross borders freely. But countries can only uphold their cultural and linguistic identity if all the relevant information is accessible and available in the national language(s). This is an important principle of the European Union today, and it also holds for all European nations. For the emergent European information society we have to develop a language technology that takes advantage of the multilingual challenge. It will have to support the production, revision, conversion, presentation, publication, documentation and last, but not least, translation of texts in technical and everyday language, and it will have to grant language-independent information retrieval by sophisticated interaction modes based on natural language. Language engineering in Europe will then play a leading role on the world market. The quality of all language technology rests on the linguistic knowledge determining the algorithms of any natural language processing application. This linguistic knowledge is accessible in and by language resources. We find it in scientifically designed text corpora, in lexicons based on existing dictionaries and on corpus analysis, and we can extract it from textual and lexical resources and convert it into the form needed in application by powerful generic software, both language specific and language independent. Language resources are the raw material of all language technology. The better they are the more expensive is their creation. The langua- ge industry, small and medium-sized enterprises in particular, often cannot afford to build them up. On the other hand, in all European countries there are focal language centres with a long tradition in the creation and application of language resources. What we need then, is a common infrastructure of (public domain) research and (private) industry. We need a common platform where providers and users of language resources come together, share expertise, discuss their needs, develop options and what is most important, exchange re- sources. In some European countries such an infrastructure exists already, in others it is gradually being built up. But most of the work was (and still is) devoted to monolingual applications. There has not been much cross-border cooperation. This is why in Western Europe several efforts have been made to build up a common infrastructure that can serve the needs of multilingual language technology applications. In March 1995, the European Language Resources Association (ELRA) was set up with strong backing by the Commission of the European Community. But Europe is larger than the European Union. Linguistic expertise, language resources and computational linguistics are highly developed in most Central and Eastern European countries. As well, we can observe here the emergence of a powerful, if still small, language industry. If we want to make Europe a competitor on the world market of language technology, we must build up a common infrastructure for the whole of Europe. The Concerted Action TRANS-EUROPEAN LANGUAGE RESOURCES INFRASTRUCTURE (TELRI), a COPERNICUS project funded by the European Commission, brings together 22 institutions of 17 European countries, with strong links to relevant language centres all over Europe. These institutions pool their resources, build up multilingual expertise, develop generic tools for multilingual applications and create a strong permanent platform for successful cooperation between research and industry. TELRI was initiated in January 1995. Already, it has succeeded in setting up several multinational joint ventures with academic and industrial partners leading to concrete language technology products for today's and tomorrow's market. These projects will be presented in the next issues of this newsletter. I strongly hope that our TELRI newsletter will make companies, organizations and institutes in research and industry aware of our activities. We need many partners with diverse backgrounds to develop a strong European network. Tell us about your needs. I am sure we will find a way to cooperate. 

WHAT DOES TELRI MEAN

TELRI partners

• University of Birmingham, School of English

Prof. dr. John M. Sinclair
Birmingham, Great Britain
Apart from the participation in TELRI, the Birmingham partners contribute to two international projects, for example EAGLES and PAROLE. As part of the EAGLES project, Birmingham is reviewing current practices in the classification of texts in major European corpus projects. On the basis of these results, the aim is for a general classification scheme, proposing a text typology suitable for European corpus work. The normal kind of text typology deals with external criteria only, very little work has as yet been done on establishing or using internal criteria as a means for identifying text types. The work will make proposals for furthering the establishment of internal criteria for text typologies. An important part of the PAROLE project is to establish links with interest groups on a national level. Birmingham has identified potential interest groups within the UK - both academic and industrial - and has met with them to discuss the possibility of collaboration and harmonisation of language resources. Such groups include universities, publishing houses, government funding bodies. A description of the PAROLE project, with links to the other partners, is available on the World Wide Web at http://clg1.bham.ac.uk/parole/ also at this address are details of a free tagging service for English texts. Birmingham is co-ordinator of the text typology subtask. The aim here is to produce specifications for the classification of texts based on the guidelines proposed by the EAGLES project. Birmingham is also co-ordinator of the subtask on spoken corpora. The aim of this report is to provide specifications for composition of spoken corpora, again based on the guidelines of the EAGLES report. Both tasks will take into account the results of the NERC final report (which is now available from our ftp server - please contact parole@clg.bham.ac.uk for details).
• Instituto di Linguistica Computazionale

Prof. Antonio Zampolli
Pisa, Italy
Main institutional goals and mandate of the Institute headed by prof. Antonio Zampolli are th following: - theoretical researches and applications in the area of computational linguistics and development of methods, techniques, instruments, basic linguistic researches, procedures. - collaboration and technical-scientific assistance to institutions and organizations that use electronic elaboration of linguistic data for scientific and application-oriented goals. - creation/distribution of resources and linguistic data and proposal/distribution of standards and generalized procedures. - development of the relationship and co-operation with international organizations working in the same area and with centres of other countries. - promotion of activities aiming at the diffusion of the scientific and technical knowledge in the area of computational linguistics.
• Charles University, Institute of Formal and Applied Linguistics

Prof. Eva Hajicova
Prague, Czech Republic
The Institute of Formal and Applied Linguistics was established in 1990. The Institute collaborates very closely with the Institute of Computational and Theoretical Linguistics at the Faculty of Philosophy, Charles University (headed by Dr. Vladimir Petkevic). Both institutes continue in the work of the former group for computational lingusitics existing at Charles University since 1959. The main research interests, resources and expertise of both institutes are as follows: - dependency syntax - morphological analysis, POS tagging, syntactic tagging, algorithmical topic-focus identification - morphological analyzers for Czech, English, lemmatization for full-text databases - collection of daily newspaper texts - procedure for automatic identification of topic and focus, explicit description of syntactic dependency relations -English-Czech, Czech-Russian machine translation
• Tartu University, Department of General Linguistics

Prof. Haldur Oim
Tartu, Estonia
Tartu Research Group In Computational Linguistics is an interdepartmental group where participate people from the Department of Estonian and The Department of Computer science. In the 80s the Group worked mainly in the frames of artificial intelligence, dealing with text understanding and dialogue modelling. In this context we also created some experimental programs of morphological and syntactic analysis of Estonian. Since the end of 80s the work of the Group has centered on creating the computer corpus of contemporary Estonian, according to the classical principles of Brown and LOB corpora. We have put together 1 million words of texts, and at present the semi-autho-matic process of corpus tagging is on the way. A program of the morphological analysis of Estonian has been developed.
• Adam Mickiewicz University, School of English

Prof. Jacek Fisiak
Poznan
The School of English was established in 1966 and is part of the Adam Mickiewicz University.
In the domain of computational linguistic analysis, the Institute
works on
English, Polish and, to a lesser extent, other European languages
• Bulgarian Academy of Sciences, Center for Informatics and Computer Technology,

Linguistic Modelling Laboratory(LML)
Prof., Dr. Elena Paskaleva
Sofia
LML was set up in July 1987 in the framework of the Center of Infirmities and Computer Technology (CICT), Bulgarian Academy of Sciences. The Center's main objective was to coordinate scientific effort within the Academy and to promote international cooperation in the area of theoretical and practical problems of the new generations of computers.
LML Research Program
At LML research concerning NL is interdisciplinary - integrating linguists, logicians and computer scientists.
Theoretical investigations concern:
- syntax and semantics of natural languages (semantics of tenses, verb frames, etc.);
- formalisms for knowledge representation (conceptual graphs, feature languages, nets and modal languages as unifying frameworks);
- logical foundations of reasoning with imperfect information;
- problems of database theory and object-oriented methodology. Practical work is concentrated on:
- development of linguistic processors, based on different linguistic theories, and for pursuing different goals, e.g. full syntactic parsing or only partial one when concerned with a practical grammar checker;
- linguistic knowledge bases of different types and scope, e.g. computer dictionaries, morphological systems, etc.;
- intelligent information-retrieval and decision support systems based on AID principles, methodology and tools, in particular for aiding the translator of technical texts, or the practicing lawyer.
LML staff currently (spring 95) consists of 9 researchers. The main research activity of LLM is the construction of the basic components of the Bulgarian Linguistic Knowledge Base envisaged for the (hopefully not so distant) future.
The creation of this linguistic knowledge base rests upon the following basic principles:
- aiming at exhaustive models of linguistic knowledge. In view of the obvious trade-off between completeness of knowledge and depth of the language level processed, the sequence of computer model realizations chosen is from the text via morphology and lexical data to syntax and semantics;
- attacking the deep language levels not through spectacular (albeit fragmentary) illustrative models, but through gradual accumulation of linguistic knowledge from real large text corpora by means of an intelligent user -oriented interface;
- attacking the surface language levels via complete computer models, e.g. a paradigmatic model of Bulgarian morphology with a very large LDB (with an envisaged scope of 60 000 entries); a system for processing and knowledge acquisition from large texts corpora obtained from desk-top publishing systems.

National Projects
- LARGE BULGARIAN LDB
- SUPERLINGUA
-CONCEPTUAL GRAPHS - A LANGUAGE FOR DOMAIN KNOWLEDGE MODELLING
-INFERENCE CONTROL INFORMATION IN THE KNOWLEDGE REPRESENTATION SYSTEMS

International Projects
- TELRI
- GLOSSER
- BILEDITA
- LATESLAV
- ELSNET GOES EAST

• Charles University, Faculty of Philosophy

Institute of the Czech National Corpus,
Prof. Frantisek Cermak
Prague, Czech Republic
In 1992, a group of linguists and mathematicians from various Czech Universities and from the Institute of the Czech Language, Academy of Sciences, Czech Republic initiated the activities of the Computational Fund of the Czech Language. Among the main objectives of the Institute, established on the background of this initiative in 1994, there is the creation of a large general corpus that should become a versatile basis for all sorts of research and applications.
• "Jozef Stefan" Institute,

Laboratory for Language and Speech Technologies
Tomaz Erjavec
Ljubljana, Slovenia
The Laboratory for Language and Speech Technologies was established in 1988 and is part of the "Jozef Stefan" Institute. The Laboratory is engaged primarily in research of Slovenian feature-based syntax and morphology, speech generation and (computational) logic.

TELRI Working Groups

• WG1 TELRI USER GROUP

Co-ordinator: Wolfgang Teubert
TELRI is aimed at promoting cooperation between public domain research institutes and language industry. Particularly small and medium-sized enterprises, software developers, publishers or translation services, cannot afford to build up their own language resources. They have to rely on affordable resources provided by TELRI partners or similar institutions. Only real projects aimed at the development of a marketable and concrete product will reveal which kinds of resources are needed. Therefore, each TELRI partner is asked to link with three companies either in a joint venture, or by supplying resources or other services to a commercial project. All TELRI partners take part in this effort, thus laying the foundation for a permanent platform where suppliers and users of language resources meet, share their knowledge, exchange data, and design new computational tools.
• WG2 DOCUMENTATION

Co-ordinator: Ruta Marcinkeviciene
The aim of WG2 is to collect, document and disseminate relevent information on 1. language resources such as corpora, lexical databases, machine readable dictionaries plus lingware 2. language engineering organizations of all possible types: providers of data, users of data or of a mixed type 3, projects, meetings, conferences and other activities in the field of language engineering 4, relevant literature. The outcome of our WG as well as all national TELRI partners' work should be a distributed knowledge database available both in an electronic and printed form. The followinf steps are to be taken in order to create the above mentioned database: - the data to be documented defined and relevant actors identified, - the requested data collected, - the database designed and implemented, - the database access mean implemented and partnets as well as other countries informed.
• WG 3 NEWSLETTER

Co-ordinator: Eva Hajicova
The main task of the working group is to prepare and publish in regular intervals (three times per year) TELRI Newsletter informing the academic community, their industrial partners and also the prospective users about the activities of individual TELRI working groups, about available resources and about methods for their processing. The Newsletter helps in this way to make the communication between all interested parties easier and more effective.
• WG4 TELRI SEMINARS

Co-ordinator: Julia Pajzs
The task of this working group is the organization of three "TELRI seminars". The first one, named "The European Seminar Language Resources for Language Technology" will take place in Tihany, Hungary 15-16 September, 1995. The aim of the seminars is to bri So far our working group succeded in collecting a mailing list for the first circular of the seminar, the location and time of the seminar was chosen. A hungarian small private service company was trusted by organizing the seminar. The first circular is to be distributed in a short time. The members of the seminar working group will meet in Hungary in July to decide on the final program and to solve any organizational problems.
• WG5 LINGWARE ASSESSMENT

Co-ordinator: Tomaz Erjavec
Working Group will assess the performance of specified lingware under controlled conditions: corpus alignment software(1995), taggers(1996), and homograph disambiguation software(1997).
• WG 6 TELRI SERVICE POOL

Co-ordinator: Pierre Lafon
Working Group will pool existing service activities of TELRI partners and develop a design for a streamlined presentation of common service actvities, including issues as charges, contracts and copyright problems.
• WG 7 TELRI NETWORKING

Co-ordinator: Vladimir Benko
Working group will build up a dedicated electronic network ( within INTERNET) between TELRI partners to enable exchange with and mutual access to each partner's language resources, and it will define standards for operation. Once in operation, this network will be open to TELRI User Group and Advisory Board members, and, by individual agreement, to other interested parties.
• WG8 LINKING

Co-ordinator: Wolfgang Teubert
This Working Group pools and documents all relevant European and international links of all TELRI partners with institutions, organisations, and projects, and establishes new links. Since some important language centres (in Croatia, Serbia and the Commonwealth of Independent States) are left out of the COPERNICUS programme this Working Group is used to have these centres participate in TELRI activities so they can become full members e Also, only seven European Union countries are represented in TELRI. It is necessary to establish links with the key centres in the other EU and European Economic Area countries to develop a common European infrastructure. For this purpose, TELRI will work c Because in Western Europe COPERNICUS activities are still felt to be only marginally important when compared to other EU activities, this Working Group must increase the EU language resources community's awareness of the available assets in Central and East
• WG9 ORGANISING JOINT RESEARCH

Co-ordinator: John Sinclair
The group will run through very modest exercises of a research nature in order to learn how to work together and share resources. One of the interesting areas for the group is that of parallel corpora, where each corpus is the translation of another, and alignment can be achieved at the sentence level. We decided to choose a classical text, on the grounds 1. that translations into most of the project languages would be likely to exist 2. that no language would be priveleged by being the original. We chose Plato's Republic. Members of the group are currently trying to find a suitable translation. If it is not in electronic form already (as the English translation by Jowett is), funds will be sought from the Co-ordinator for the cost of scanning or In the second phase of the project, the "tools" group intends to work on alignment software, so we hope to be able to take advantage of that work.
• WG 10 USER NEEDS

Co-ordinator: Andrej Spektors
The work of WG10 is aimed at the establishing of the user needs, identifying of the existing resources, preparing of the proposals for a repository of lexical resources (such as corpora, machine readable dictionaries and lexical data basis) and the developing of lingware. Every member of the group works within his country with the potential users he choosed. Afterwards the summarizing of the results will be made and the common features will be generalized.
• WG 11 PERMANENT INFRASTRUCTURE

Co-ordinator: Antonio Zampolli
Working Group will prepare a proposal for the design of a permanent infrastructure to extend and complement the infrastructure that will be set up for the European Union language industry.

TELRI Events - 1995

• JANUARY 13 &14:

TELRI Inaugural Meeting, Mannheim / Germany Plenary Meeting, Steering Committee meeting and meetings of TELRI working groups
• FEBRUARY 28:

Meeting of the ELRA ( European Language Resources Association) Interim Steering Committee, Luxembourg / Luxembourg. Wolfgang Teubert takes part in the meeting. As a member of this committee he furthers the interests of the Central and Eastern European countries.
• MARCH 27:

Meeting of the ELRA Interim Steering Committee, Brussels / Belgium. Wolfgang Teubert takes part in this meeting.
• APRIL 3-16:

Visit to China by Wolfgang Teubert as co-ordinator of TELRI Working Group "Linking" to form contacts with promising language centres in China (Institute of Applied Linguistics, part of the Chinese State Language Commission in Beijing: Prof. Feng Zhiwei / Institute of Natural Language Processing of the Jiao Tong University, Shangai: Prof. Yang Huizhong / Departament of Computer Science of the Tongji University, Shangai: Prof. Sheng Huanye and Prof. Wu Quidi). Results are conceptions of some low-scalejoint projects that could lead to medium-term cooperation between TELRI members and Chinese partners.
• MAY 15:

Meeting of the ELRA Steering Committee, Brussels / Belgium. Wolfgang Teubert takes part in the meeting.
• MAY 18-21:

Visit to Belgrade by Wolfgang Teubert as co-ordinator of Working Group "Linking". The visit to Belgrade led to concrete plans for the participation of the Chair of Computer Science of the Faculty of Mathematics at Belgrade University, Prof. Dusko Vitas. The goal is - even though Serbia, Croatia, Bosnia and Macedonia are excluded from the COPERNICUS programme - to use existing contacts to link key institutions there to the TELRI Concerted Action.
• MAY 27:

TELRI Steering Committee Meeting, Florence / Italy
• MAY 27:

Meeting of TELRI Working Group 2 "Documentation", Florence / Italy
• MAY 27:

Meeting of TELRI Working Group 8 "Permanent Infrastructure", Florence / Italy
• JUNE 10:

Meeting of TELRI Working Group 7 "Networking", Bratislava / Slovakia On the second weekend of June, the 2nd TELRI WG7 meeting (originally scheduled for April) took place in Bratislava, Slovakia. All the WG7 member sites were represented, mostly, by the "instead-of's": Oliver Jakobs (BIR), Cyril Belica (MAN), prof. Janusz Bien (WAR), and Dr. Eduard Kostolansky and Vladimir Benko (BRA1). Tomaz Erjavec (LJU2), the WG5 coordinator, was also present at the meeting. The topics discussed covered experience with electronic communication among the TELRI partners in general, the current state of the facilities provided (TELRI mailing list established in Mannheim, TELRI WWW home page currently maintained by Tomaz Erjavec in Ljubljana and Edinburgh), specific problems of "slow" TELRI sites and the question of using national character sets in e-mail communication and WWW access. It has been agreed that the TELRI WWW home page is to be moved to Mannheim where it will be maintained and updated on a regular basis. The "Internet How-To Manual" for TELRI partners that is being prepared will concentrate mostly on basic procedures and suitable "entry points" where additional information is to be found.
• JUNE 23 & 24:

Meeting of TELRI Working Group 3 "Newsletter", Prague / Czech Republic
• SEPTEMBER 11-14: TELRI Plenary Meeting, Steering Committee meeting, meetings of all TELRI Working Groups, Tihany / Hungary
• SEPTEMBER 15 & 16:

TELRI Seminar, Tihany / Hungary

Top of this issue

TELRI Main Page