ELSNews, vol. 4.6, December 1995

Permission is given by the Editor to distribute or re-use ELSNews articles appearing in the paper version of the newsletter, or in these Web pages, with the proviso that the following text is included with the re-used article:

``This article appeared in ELSNews 4.6 (December 1995), and is re-printed by permission from the Editor. ELSNews is the newsletter of ELSNET, the European Network in Language and Speech. Information about ELSNET is available from the Coordinator, at elsnet@let.ruu.nl.''

Table of Contents

ELSNET Goes East
Reports on the Moscow workshop and the Russian Network Awareness Day
Application-oriented research
The HARP system provides speech training for hearing-imparied people
1996 Summer School on Dialogue Systems
First announcement of ELSNET's annual summer school
Text Encoding Initiative
What is the TEI?
Future Events and Misc
Return to the ELSNET home page.

Looking Back, Looking Forward

Steven Krauwer, ELSNET Coordinator, Utrecht University

Now that 1995 has come to an end, it seems appropriate to look back on our achievements of the past four years, and to say some words on the period ahead of us.

ELSNET began in 1991 with 25 academic nodes. In the last four years its membership has increased to about 100 nodes---60 academic and 40 industrial --- from nearly every state in Western Europe and even a few in Eastern Europe. One important achievement was the writing of the report, Strategic Research in Speech and Natural Language. This document outlined a long-term research strategy that advocated an evaluation-based approach to research, having the advantages of schemes like DARPA in the US, without having its disadvantages. Although we can't be sure about the causal effects of the report, it was clearly an early contribution to a growing interest in evaluation that can now be discerned in many recent and ongoing EC projects. In the area of training, the annual European Summer School has been very successful in attracting broad participation from academia and industry. Training and mobility of researchers in ELSNET has been implemented via the HCM programme. Links between industry and academia have been established and ELSNET's information dissemination structures have proven to be powerful instruments for ensuring a continuous bi-directional flow of information between academia and industry and between language and speech communities. In the area of resources ELSNET has, through its spin-off project, RELATOR, been instrumental in the creation of ELRA, the European Language Resources Association. CD-ROMs with language resources have been produced and distributed. Important initiatives have been taken with respect to Eastern and Central Europe, such as the Survey of Language Engineering Organisations in Eastern and Central Europe, and the Copernicus project, ELSNET goes East.

We can conclude from all this that ELSNET is playing a significant role on the NLP and Speech scene in Europe. This seems the right time to thank all who have contributed so much to ELSNET's activities in the past --- Ewan Klein, my predecessor as ELSNET's coordinator, his assistant Dawn Griesbach, the members of the Executive Board, the convenors and members of the Task Groups, and many others in the ELSNET community. I hope that we may continue to rely on their (and everybody else's) support in the future. But what will the future bring? As many of you know, a proposal has been submitted for the con-tinuation of ELSNET, under the title ELSNET-2. This proposal is now under review, and an unfunded extension of the current contract will allow us to continue our activities until July 1996. Below is a brief sketch of the next phase of ELSNET.

ELSNET-2 will focus on four areas: integration of language and speech; integration and comparative evaluation of research results; the relationship between academia and industry; and language coverage. Workshops, with participation from academia and industry, will be held on hot research topics. Proposals --- to be prepared and submitted for funding from other sources --- will aim to develop new evaluation methods for integrated systems, and to create a European evaluation infrastructure. The annual Summer School will continue, common curricula in NLP and speech will be developed, and facilities will be offered for student placements in industry. In the area of resources, ELSNET will work closely with other related European actions. Links between academia and industry will be strengthened through intensified dissemination of information and by the creation of an Industrial Panel which will be consulted regularly. The establishment of ELSNET as a legal entity will give the network more visibility, and allow it to generate its own income, making it less reliant on EC funding. Finally, ELSNET-2 will continue to open up towards the East, with the medium-term goal to create a single pan-European Network in Language and Speech.

Some who commented on draft versions of our proposal said that our plans were rather ambitious in comparison to the financial contribution requested from the Commission. This is true. But we must realise that we are a network, not a project, and that our main resources are not primarily the ECUs from the Commission, but rather the commitment and the enthusiasm of our members. I don't know how many individuals can be counted as members of the ELSNET community. With 100 nodes having an average of 15 NLP/speech people each, this would amount to 1500 people. If everyone is prepared to spend a single working day per year as an active ELSNET member (e.g., member of a committee, organiser of an event, collector of information, author of a report or proposal, teacher at a Summer School, participant in a workshop, etc.), this would result in some 7.5 person years available to the Network per year, on top of the 2.5 person years paid for by the Commission. In ECUs (at a conservative rate of 700 ECU per day) this would be a contribution from the members of more than one million ECU!

I wish you all a very Happy New Year, and I look forward to our future collaboration in order that we may get a bit closer to our common goal.

ELSNET Goes East . . . and stays there

Erik-Jan van der Linden, University of Amsterdam, project coordinator of ``ELSNET goes East''

``ELSNET goes East . . . and stays there.'' That was what one of the participants in ELSNET's Moscow workshop said. It is a slogan that aptly expresses the way in which European R&D cooperation is becoming more and more ``Pan-European.'' The reports (given below) on the workshop and on the `Russian Internet Awareness Day' that preceded it convey this impression as well.

In general, Central and Eastern European researchers working in industrial and academic language and speech institutions are very much aware of the importance of the Internet in establishing a pan-European research community. A lot of energy is being invested in overcoming the current technical problems. Inexperienced users are eager to learn about the Internet, and a number of researchers are working on language and speech services for the Internet itself. As a result of all this activity, various small enterprises in language engineering are coming into existence in Russia. ELSNET goes East is gathering information about such companies, and hopes to make this information available in the near future through WWW-services it will develop in cooperation with the Russian R&D company, ANALIT, Ltd. Updates on this work will be given in future issues of ELSNews.

Networking Awareness Day Held for the Russian Language and Speech Community

Vera Semenova, ANALIT, Ltd.

A one-day open seminar took place in Moscow, on November 6, at the Russian State University of Humanities (RSUH), in order to bring together researchers and providers in Russia's language engineering community. The seminar was divided into three main sessions, each lasting between 1.5 and 2 hours. The day was organised by the Moscow R&D company, ANALIT, Ltd. in collaboration with RSUH's Faculty of Theoretical and Applied Linguistics (FTaAL), the Institute of New Technologies in Education, and ELSNET Goes East. While most of audience came from Moscow, and towns around Moscow, there were also a number of attendees from St. Petersburg, Novgorod, Tver, and even as far away as Kiev in Ukraine, and Vladivostok in the far East.

The first session, called ``What is Internet?'' began with a presentation by Olga Galkina and William Fick, of the US-based International Research & Exchanges Board (IREX). They gave an ``Introduction to Internet'' and presented a document entitled Frequently Asked Questions: E-mail and Internet in the NIS and Baltics (see p. 5). The second talk --- ``Internet for the Language and Speech Community'' --- was given by Erik-Jan van der Linden (ELSNET, University of Amsterdam). The final presentation in this session, given by staff from the Moscow Education Department, was called ``Internet: New Opportunities in Education.''

The second session of the day focused on the topic of ``Networks in Russia,'' and was begun by Alexey Platonov (director of the Russian Institute for Public Networks) who talked about ``The Internet in Russia.'' He spoke in a general way about the situation in Russia, describing some of the achievements and obstacles in the development of public networks in that country. This talk was followed by an on-line demonstration which showed the audience where the home pages of Russia's network providers and lingware producers could be found. (Because of problems with poor quality phone lines in Russia, this demonstration had to be very carefully prepared in advance.) The presentation resulted in a lively discussion in which some of the members of the audience shared their knowledge of networks and services in Russia. Andrey Sebrant, the manager of the GlasWeb project (which in engaged in WWW server development at GlasNet, a Russian network), Egor Anoshkin from Paragraph International, one of Russia's major producers of lingware products, Valerian Khutoretsky, head of the Moscow Information Center RAS-STN, and Alexey Platonov were the main participants in this discussion.

The ``IBM Open Net in Russia'' was presented by Eugene Ravich from IBM East Europe/Asia, Ltd. (a Russian subsidiary of IBM). Dr. Victor Veselago, head of INFOMAG, spoke next. INFOMAG is an Internet service which provides, among other things, free bibliographic resource retrieval and distribution. Marina Karelina, expert at the Russian Law Academy, explained some of Russia's laws and bills in the informatics area and then described some of the legal problems associated with the development of Russian networks.

The third session of the seminar considered ``Russian Language Engineering: its infrastructure and the use of networks for its development.'' This session was opened by Alexander Barulin, the Dean of FTaAL, who gave a talk entitled, ``Informatics in Russian Humanities.''

This was followed by a series of short presentations on:

  • ``The training of future linguists at the University of Humanities'' (Alexander Barulin, FTaAL Dean);

  • ``The DIALOGUE conference'' (Alexander Narin'yani and Sergey Sharov, Russian Institute of AI). (DIALOGUE 95 was an international revival of a workshop for computational linguists held regularly in Russia in the years between 1975--1988.);

  • ``The new conception of the Computerized Fund of Russian Language (CFRL)'' (Vladislav Andryuschenko, Director of CFRL at the Russian Language Institute, Russian Academy of Sciences);

  • ``Survey of Russian teams and products in language engineering'' (Vera Semenova, General director of ANALIT, Ltd);

  • ``The ELSNET goes East project'' (Erik-Jan van der Linden, project coordinator and Igor Boguslavsky, head of the Russian coordinating node).

    The seminar ended with a well-received presentation by Poul Andersen (European Commission, DG XIII) on the topic of ``Trans-European cooperation with Russia in the LE-area: Today and in the Future.'' This talk gave a very helpful overview of European programmes that aim to encourage cooperation on the pan-European level, and in particular, those in which Russia may participate.

    In order to get a better idea of the composition and background of the audience attending the seminar, the organisers distributed a questionnaire. Of the 142 questionnaires given out, 103 were completed and returned.

    It was found that 47% of the respondents represented R&D institutes, 33% were from universities, and 25% worked for private industrial companies. There were also a few individuals attending from international organisations, libraries, schools, mass-media, etc. (Because the answers to individual questions were not mutually exclusive, it is sometimes the case that the response percentages for some questions might total more than 100%).

    Not surprisingly among such an audience, fewer than 10% of the participants had never used linguistic technologies, whereas more than 80% used them regularly, with information retrieval from text databases being the linguistic technology used most commonly (i.e., by 44% of the respondents). Other technologies include: electronic dictionaries (used by 43%), spell checkers (used by 34%), machine translation systems (used by 24%), linguistic processors (used by 19%), and OCR tools for text input (used by 18%). Forty-eight percent (48%) of the participants were directly involved in the development of linguistic technologies.

    Most of the respondents were users of Russian networks, with 39% using Relcom; 8% using GlasNet; 6% using Sovam; and 4% using Demos. Twenty-three percent (23%) had occasionally used networks, while only 12% had never used networks before.

    Due to the events of the last 10 years, the infrastructure which supports the Russian language and speech community has been almost completely lost. Events like the network awareness day are crucial for the re-establishment of this infrastructure and for the strengthening of the research community.

    ANALIT, Ltd., in collaboration with CEA (France), has published a catalogue of Russian teams and resources in language engineering on behalf of the French Ministère de l'Enseignement Supérieur et de la Recherche. A revised edition of this catalogue is soon to be installed on an Internet WWW-server. More information about the catalogue is available from:

    Vera Semenova
    ANALIT General Director
    7-2-429 Sayanskaya St.
    Moscow 111531, Russia
    Fax: +7 095 307 2768
    Email: vera@analit.msk.ru

    Workshop on the `Integration of Language and Speech: Academia and Industry'

    Ryszard Gubrynowicz, Speech Acoustics Laboratory, Institute of Fundamental Technological Research, Polish Academy of Sciences

    Well, it was great for `ELSNET Goes East' to go as far east as Moscow, although the weather was a bit cold with the statutory snow there at this time of year. The Research Workshop on Integration of Language and Speech was held on Nov. 9-10,1995 in Domodedovo, a new town around 40 km outside Moscow. This workshop was one of the planned activities within the COPERNICUS project `ELSNET Goes East', and it had the goal of contributing to the extension of ELSNET beyond the borders of Western Europe. The meeting provided an overall picture of the language and speech engineering profession in Central and East European countries (including the Newly Independent States of the former Soviet Union). It was also a forum for discussion between language and speech researchers, working in different environments, e.g., research institutes, universities and industry, and in different countries.

    The Institute for Information Transmission Problems (IPPI) of the Russian Academy of Sciences was the host of the workshop. Apart from the main financial support from `ELSNET goes East,' a supplementary contribution came from the Ministry of Science of Russia and Russian Foundation of Basic Research. The organization committee was headed by Igor Boguslavsky who was assisted very effectively by his colleagues from the IPPI. After opening talks by Nikolai Kuznetsov (IPPI Director), Steven Krauwer (ELSNET coordinator), Poul Andersen (EC), and E.-J. van der Linden (`ELSNET goes East' coordinator), the workshop started with presentations of the current state of the art in machine translation (S. Krauwer, Utrecht University, The Netherlands) and in the creation, management and exchange of language resources. Wolfgang Teubert (Institut für deutsche Sprache, Germany) presented information about the Concerted Action, TELRI, which is devoted to international cooperation in the creation of a language resources infrastructure, where providers and users of language resources come together from the European Union as well from C&EE, discuss their needs, and exchange their resources. Joseph Mariani (LIMSI-CNRS, France) gave a survey of existing language resources with an evaluation of each. A very interesting survey of Russian teams and products in language engineering was presented by Vera Semenova (ANALIT, Ltd, Russia).

    During the two days of the workshop, several theoretical and practical aspects of integration of Natural Language Processing and Speech Processing were discussed. From the linguistic point of view, Olga Krivnova (Lomonosov State University of Moscow, Russia) and Eva Hajicova (Charles Univ, Czech Republic) spoke on the topic of prosody. The speech researchers, Juri Kosarev (St. Petersburg Institute for Informatics and Automation, Russia) and Taras Vintsiuk (Glushkov Institute of Cybernetics, Ukraine) spoke about the integration of linguistic knowledge in speech processing and recognition. Igor Boguslavsky, Boris Lobanov, and Elena Kornevskaya (both from the Institute of Engineering Cybernetics, Belarus) described an on-going collaborative project, involving both language and speech teams, which aims at the incorporation of morpho-syntactic knowledge into the intonation and accentuation processor of a speech synthesizer . The presentation provided a very convincing example of how cooperation between the fields of language and speech could be effective on the common technological grounds. Some specific problems related to speech processing like phonetic transcription, speech pronunciation, production, recognition and perception were presented. Two papers were devoted to speaker identification. Among NLP topics one presentation considered two approaches to text analysis. One of industrial representatives, Jean-Pierre Chanod, presented a large survey of the research and products of Rank Xerox Research Center in Grenoble.

    After the conference, on Saturday in the morning, the organizers arranged a brief visit to the Kremlin and a very colorful Kolomenskoye monastery. In the afternoon, after having lunch in an Italian pizzeria, the participants visited the Laboratory of Computer Linguistics of IPPI, where an impressive bi-directional Russian-English machine translation system was presented.

    The atmosphere during the workshop was very pleasant and helped to achieve the goal of re-establishing better scientific ties between the linguistic and speech community working in C&EE. The workshop also provided an opportunity for people coming from Western Europe and C&EE countries to get together and make arrangements for future collaboration. Forty participants coming from 11 countries contributed to make the workshop a European event. The proceedings of the workshop, which are under preparation, will contribute to information dissemination on a multi-national level and strengthening the European research community in the area of integrated NLP and SP research.

    The coordinates of the author are:

    Ryszard Gubrynowicz
    Speech Acoustics Laboratory
    Institute of Fundamental Technological Research
    Polish Academy of Sciences
    Swietokrzyska 21, st.
    00-049 Warszawa, Poland
    Tel: +48 22 261 281
    Fax: +48 22 269 815
    Email: rgubryn@ippt.gov.pl

    For information on the workshop proceedings:

    Igor Boguslavsky
    Laboratory of Computer Linguistics
    Institute for Information Transmission Problems
    Russian Academy of Sciences
    Bolshoy Karetny, 19
    Moscow, 101447, Russia
    Tel: +7 095 209 4927
    Fax: +7 095 209 0579
    Email: bogus@ippi.ac.msk.su

    Behind the scenes of the Moscow workshop

    Ingrid van Loon, University of Amsterdam

    Having been the new project administrator for ELSNET goes East for only six weeks, I had the pleasure to go to Moscow. The reason for my visit was the Workshop on Integration of Language and Speech, which was held on Nov. 9-10. The workshop was combined with a meeting of the Steering Committee of ELSNET goes East on Wednesday. The organisation of the Workshop was in the hands of Igor Boguslavsky of the Institute for Information Transmission Problems (Russian Academy of Sciences). He was assisted by Alexandre Lazourski, Irina Kajali and Natalia Ikoeva. These four formed an efficient---and above all---warm and very friendly team of hosts.

    The workshop and the meeting of the Steering Committee took place in the boarding house Elochki, in the far outskirts of Moscow. This multi-functional place also accommodated all the participants, and provided three hot meals per day for them. At the beginning it was a bit strange to begin your day with noodles or rice, but after some time, you get used to this. On Wednesday morning the meeting of the Steering Committee took place. In general, it was concluded during this meeting that the climate for cooperation between Western and Eastern Europe is improving. ELSNET is moving in the direction of a Pan-European network. Another vital development is the growing use of Internet in Central and Eastern Europe. This had become apparent at the Muscovite congress called `The Internet Awareness Day' on November 6, in which ELSNET goes East's project manager, Erik-Jan van der Linden, delivered a presentation on ``Internet for the Language and Speech Community.'' A number of other subjects were discussed in the Steering Committee meeting: the coming workshop on the integration of Language and Speech, the Summer School in Budapest in July, the cooperation between ELSNET goes East and TELRI (another Copernicus project at the field of Language Resources), and the current activities of ELSNET.

    In the afternoon a visit to the Institute for Russian Language (which is part of the Russian Academy of Sciences) was included in the schedule. Researchers from the lab gave us several demonstrations. I have to admit that, having studied Economics and Sociology of Third World Countries, I did not understand much of it. What I did understand was that everybody was impressed by the great amount of interesting work that is being done with so little money and so few means. The conditions under which the people are working are really hard. Often they have little more than the bare necessities in terms of housing and equipment for carrying out research. People do not know whether their institute will still exist next year. For most Western European readers of ELSNews, this will be hard to understand.

    On Thursday and Friday the workshop took place. A scientific report on the meeting is given on pages 3-4, by Ryszard Gubrynowicz. I'll continue with what was for me the best part of the workshop: the banquet on Thursday evening. With all forty workshop participants gathered around one long table, we feasted on a variety of Russian dishes, including caviar. The dinner was interrupted several times by speeches and words of thanks to the organisers of the event. Starting himself, the Belarussian Boris Lobanov provoked the participants to sing a song of his or her own country. Of course everybody accepted this challenge and tried to sing better and louder than those who had gone before.

    On Saturday an excursion was organised by Igor Boguslavsky and his colleagues for those participants who were staying over the weekend. They arranged a very good guide, who had a great way of telling stories. We visited the great museum, Oruzhejnaja palata (The Armoury), in which all the treasures of the Russian empires were stored. For an hour-and-a-half our guide told us gossip stories about all the tsars, without boring us for a second.

    After having lunch at a pizzeria, we were invited to see the lab of Igor Boguslavsky. He showed us a machine translation system, which looked very impressive to me. My layman's judgment was confirmed by people who had real knowledge of it: everybody was again very laudatory about the potential of the system. On Sunday the remaining participants left for the airport.

    Back in Holland I tried to evaluate my time in Moscow. What made my stay so worthwhile was the social part of it. I had the chance to get to know people personally whom I had known previously only via e-mail contacts. I think it was very useful for our future cooperation. Besides, Moscow is a great city and it was a pity that we had so little time to visit. Finally, the organisational talents of Igor Boguslavsky and his colleagues made the workshop a very successful event. As Erik-Jan van der Linden stated: ``Nothing in the participants perception of the workshop organisation must be taken for granted.'' The team of organisational `wizards', as we called them, worked very hard to realise this workshop. Thanks, thanks, thanks!

    Details about the activities of ELSNET Goes East may be received from:

    Inst. for Logic, Language & Computation
    University of Amsterdam
    Plantage Muidergracht 24
    1018 TV Amsterdam
    The Netherlands
    Tel: +31 20 525 6051
    Fax: +31 20 525 5206
    Email: ingridvl@fwi.uva.nl

    IREX Answers FAQs

    The International Research and Exchanges Board (IREX) is a non-profit organisation that encourages intellectual exchange between the US and the countries of the former Soviet Union, Central and Eastern Europe, and Mongolia. The organisation has published a document entitled, Frequently Asked Questions (FAQ): E-mail and Internet in the NIS and Baltics. This 24-page booklet tells you, in simple, straightforward language, everything you ever wanted to know about computer communications facilities in the Newly Independent States (NIS) and the Baltic States. Its goal is to attempt to answer the questions most often asked by scholars travelling to this part of the world, and by those seeking to help NIS and Baltic colleagues join computer networks. It answers questions like:

    The booklet provides comprehensive information about each of the most well-known Internet service providers in the Baltics and the NIS, including contact names and addresses. It also gives a list of universities and Academy of Sciences institutes which have developed their own network host systems.

    The document is available from IREX, or via the worldwide web by looking here. A Russian version of the document is available from IREX's Moscow office.

    Further information about IREX and its activities is available from:

    1616 H Street, NW
    Washington, DC 20006, USA
    Fax: +1 202 628 8189
    Email: irex@info.irex.org

    The Moscow office may be contacted at:

    14/5 Volkhonka 5 Fl
    Moscow 119842
    Tel: +7 095 203 9889
    Email: open@glas.apc.org

    Application-Oriented Research

    HARP: Speech Training for Hearing-impaired People

    Edmund Rooney and Fabrizio Carraro, Centre for Communication Interface Research, University of Edinburgh, and Nikki Pattinson, Future Speech Systems, Ltd.

    HARP is a two-year EC TIDE project which is developing a speech training aid for hearing-impaired people. The aim is to develop a flexible, relatively inexpensive teaching tool, based on a standard multimedia PC, which will be used to complement and extend the services offered by speech therapists and teachers of the deaf to their hearing-impaired clients. The system provides real-time visual feedback on parameters such as voice fundamental frequency, intensity, spectral quality and timing, for both English and French speakers, using the latest developments in multimedia and a range of speech analysis facilities.

    The HARP consortium comprises one academic and two industrial partners: the Centre for Communication Interface Research (CCIR) at the University of Edinburgh; Future Speech Systems, Ltd. of Lanark, Scotland; and Agora Conseil of Grenoble, France. Future Speech Systems (FS2) is an SME specialising in the exploitation of speech technology for the disabled, while Agora specialises in consultancy and technology transfer. The consortium's sponsoring partners, who are providing advice and support in the design of the system, as well as access to user groups for the trial and evaluation of the system, include the otolaryngology department of the Royal Infirmary in Edinburgh, Donaldson's College for the Deaf in Edinburgh, and the Institut des Jeunes Sourds in Chambéry, France. Valuable assistance has also been received from staff at the Department of Speech and Language Sciences at Queen Margaret College, Edinburgh.

    Speech training aids for the hearing-impaired

    The use of computers to give visual feedback to deaf speakers has been the subject of research for many years. The aims of this research have been to make speech therapy more effective, not only by providing objective feedback to supplement the efforts of the therapist, but also by freeing the therapist from time-consuming practice sessions, and giving the hearing-impaired users greater opportunities for practice using stimulating and motivational material.

    A range of systems has been developed, using both acoustic and physiologically-based measures of speech performance. Acoustic systems, using only a microphone for input, have the advantage of simplicity of operation and relatively low cost, but are unable to give feedback on many aspects of production, such as nasality and tongue position. Physiologically-based systems overcome this limitation by supplementing the acoustic information with more sophisticated measurements such as airflow, laryngeal behaviour and electropalatography, but the equipment used to obtain these measurements is often delicate, prohibitively expensive and difficult to use without extensive training.

    Despite the considerable research effort devoted to this area, very few of these training aids have become widely available commercially. Those that have use only acoustic measures, since this technology is relatively simple, cheap and non-invasive, but they are all still relatively expensive, and require either dedicated PCs or specialised hardware units which are added on to a standard computer. These considerations have hindered the widespread acceptance of such systems into schools and clinics, and ruled out the provision of additional units for use in the home. Inadequate and demotivating feedback is another problem, particularly for those systems which use whole word distances or overall ``goodness'' scores: users of these systems may be unable to judge exactly which parts of their utterance are unacceptable, and may not be able to interpret the displays in a way which allows them to modify their production to move it nearer to the intended target. These problems mean that many of the aims of computer-based teaching are not being met, since users either have restricted access to systems, or find that they cannot use the system without constant supervision by the therapist.

    Design specifications for the HARP system

    The HARP system represents an attempt to address some of these problems using the latest developments in speech processing and PC design to produce an accessible, robust and motivating speech training tool which may be used in hospitals, clinics, schools and the home. Like the existing commercial systems, the HARP system is acoustically based for ease of use and maximum accessibility, requiring only a microphone for speech input. Unlike other systems, however, HARP is based on a standard multimedia PC (running Windows), with speech input and output performed on a Soundblaster board, now widely established as the standard for PC-based speech applications; no additional (and expensive) speech processing hardware is required.

    The system is being developed for both children and adult speakers, with four main user groups in view: speakers with early or congenital hearing loss, who have never had the benefit of normal auditory feedback and whose speech is typically highly distorted as a result (the pre-lingually deaf); speakers whose hearing loss has occurred after the onset of normal speech (the post-lingually deaf); elderly speakers experiencing problems with loudness and speech timing; and cochlear implant users, who need training to interpret the feedback provided by their implant. The system is being designed as an aid to therapists during therapy sessions with their clients, but with extensive provision for autonomous use by clients to supplement this supervised training and improve the prospects for the ``carryover'' from their lessons into their everyday speech.

    One of the key design features of the system is its use of multimedia technology. To cater for the wide range of users with greatly differing linguistic and physical abilities, for example, the system accepts input using the keyboard, the mouse or a touch-sensitive screen. Output from the system uses the multi-media facilities of the host PC, including speech, graphics and video playback. In the pitch module, for example, the user controls the playback of a set of short video sequences, their pitch governing either the speed or the direction of playback of the sequence. Other modules use simple but motivating computer graphics, in the form of a range of voice-controlled games.

    A full multimedia HELP system has also been designed and implemented. This HELP system provides both text-based information on window content and button functions, for example, and video-based information on the use of the system and the activities available: users can choose subtitled speech in English or French (see Figure 1) or the corresponding sign languages (British Sign Language and Langue des Signes Française) (see Figure 2).

    A second factor in ensuring the effectiveness of the system is the range of speech analysis facilities being provided, using speech technology originally developed within the SPELL project (Hiller et al. 1993) for teaching pronunciation to foreign language students. These facilities have been extended to include a number of analysis modules providing real-time feedback, with real-time fundamental frequency measurement using a time domain pitch tracker, real-time amplitude measurement (given by the same algorithm) for use within various loudness teaching modules, continuous vowel quality assessment employing real-time LPC-based formant analysis (a modified version of the McCandless algorithm (McCandless 1974)), and fricative assessment using measures of spectral energy distribution. The provision of continuous, real-time feedback by the system allows users to monitor their performance using kinaesthetic or tactile feedback as they watch the visual cues on the screen, giving them immediate and positive reinforcement.

    For other aspects of speech performance and language use, the HARP system incorporates a hidden Markov Model automatic segmenter, which is tuned to the speech characteristics of the hearing-impaired, allowing it to monitor the segmental content of the input. This allows the use of real language materials such as words and short phrases during training, but still permits the system to give feedback on the pronunciation of individual segments where required. It also gives the system greater robustness in the face of gross pronunciation errors or mistakes by the user, and makes unsupervised use a more practical possibility.

    HARP teaching modules

    A major requirement for the success of a speech training aid is the provision of a body of motivating courseware which allows speakers to learn from the feedback they receive. The system needs to maintain speakers' interest and avoid the frustration caused by the use of repetitive or limited materials. The HARP system therefore provides a range of courseware modules, starting with the acquisition of basic speech skills such as pitch and loudness control, and progressing to the control of intonational and rhythmic features in connected speech. This has involved the development of a variety of graphical approaches to feedback including contour-based displays, two-dimensional charts and entertaining games.

    The modules which have been developed so far cover four general areas: pitch, amplitude, consonant production, and vowel quality. The modules provide simple, reinforcing feedback to make users aware of variations in these parameters, and offer easily achievable targets as one of the first steps in rehabilitation. The Pitch teaching modules, for example, are designed to develop an awareness of pitch variation in speech and the ability to match simple phonetic targets with pitch. The activities make use both of simple computer animation and of the multi-media video facilities incorporated into the system (e.g., the user's pitch determines the speed of playback of a short video clip, high pitch being equated with rapid movement on the screen). Similarly, the Amplitude modules develop an aware-ness of loudness variation and an acceptable loudness range; loudness is equated with a number of physical dimensions, including the size, the intensity and the proximity of objects.

    The Consonant teaching modules offer both continuous feedback, to teach features such as the production of fricatives at different places of articulation, and categorical feedback using whole word utterances to teach segmental contrasts or phonetic features such as the difference between stop and fricative articulations.

    In the module for Vowel Quality teaching, users receive feedback on their own vowel production in comparison with a suitable target, and also on the approximate location of the categories corresponding to the vowel phonemes of English and French. This is done by providing a two-dimensional chart which approximately represents the articulatory dimensions of tongue height and tongue frontness/backness. A set of vowel targets superimposed on this chart allows users to monitor their own vowel quality as their articulation changes, and to compare the quality they are achieving with these target locations. (See Figure 3 on p. 8).

    Teaching modules for intonation and rhythm are also being developed. All the courseware modules have been implemented using a mixture of Microsoft Visual Basic and Microsoft Visual C++.

    Evaluations and User Trials

    In order to ensure that the design of the system reflects the needs of users, the prototype HARP system is being evaluated on a continuing basis, with the help of hearing-impaired speakers and therapists in Britain and France. These evaluations have been divided into three stages: i) a preliminary stage involving extensive consultations with user groups; ii) a period of formal evaluation of the prototype; and iii) a set of controlled Performance Trials of the system with all four user groups.

    The first two of these stages have now been completed, with encouraging results. In particular, the evaluations of the prototype system, although exploratory in nature, have been very promising. The goal at this stage was to let therapists use the system freely to supplement their normal teaching activities, in order to gather information about the system and to implement improvements in response to users' requests. The results of these evaluations have confirmed the benefits of providing real-time feedback with the power of a multimedia system, and yielded useful suggestions for improving the user interface and courseware, many of which have now been implemented.

    The evaluations have also revealed a significant demand for visual feedback systems like HARP in the treatment of other speech disorders, such as dysarthria, aphasia and phonological impairments. Therapists at one site have begun using some of the existing modules in their work with a dysarthric speaker (without hearing impairment), with some promising results.

    The final stage of controlled Performance Trials is currently underway, with the system under test in a number of speech therapy clinics in the UK and France. Initial results are expected early in 1996.

    Prospects for exploitation and development

    Throughout the development of the HARP system, the consortium has been conscious of the need to focus its attention on the future exploitation of HARP, and the eventual appearance of a market product. Work on the technical aspects of system development, described in detail above, has been matched by significant effort in extensive market research and competitive analysis by the industrial members of the consortium, who are themselves SMEs involved in the exploitation of speech technology. This research indicates that there is a large market for a well-designed, robust training tool which can offer therapists and hearing-impaired users the flexibility and control they desire.

    The consortium has also worked hard to ensure that users have been closely involved at all stages, from the initial specification to the trials. As part of this effort, the HARP Interest Group was established mid-way through the project, comprising speech therapists, teachers, hearing-impaired users from many countries and a range of organisations, all of whom have expressed their interest in a market product. This Interest Group is helping to provide valuable information on product specification, pricing and user support, in addition to the assistance with development and design coming from the Focus User Groups provided by the project's sponsoring partners and other institutions.

    Technical developments in the system have been undertaken with a view to minimising the risks associated with their exploitation. Future development costs have been kept relatively low by the choice of existing technology, with no requirement for special hardware, and by a careful adherence to the emerging software and hardware standards. The establishment of a feedback loop from the market place to the technical team has enabled the consortium to focus its developments on what the market --- that is, the hearing-impaired users and their therapists --- actually requires. All of these factors mean that HARP is well-placed to provide a strong and viable commercial opportunity for an SME currently operating in a similar market place.


    S. Hiller, E. Rooney, J. Laver, and M. Jack (1993) SPELL: an automated system for computer-aided pronunciation teaching. Speech Communication, 13, 463-473.

    S.S. McCandless (1974) An algorithm for automatic formant extraction using linear predictive spectra. IEEE Trans. ASSP, 22, 135-141.

    F. McInnes, F. Carraro, S.M. Hiller, and E.J. Rooney (1992) Evaluation and optimisation of a segmenter for a PC-based pronunciation teaching system. Proc. Institute of Acoustics, 14, 109-116.

    E. Rooney, F. Carraro, W. Dempsey, K. Robertson, R. Vaughan, and M. Jack (1994) HARP: an autonomous speech rehabilitation system for hearing-impaired people. In Proc. 1994 International Conference on Spoken Language Processing (ICSLP94), 2019-2022.

    This article is based on a paper given at the 2nd Language Engineering Convention in London, October 16-18, 1995. Further details about the HARP system is available from:

    Edmund Rooney
    University of Edinburgh
    80 South Bridge
    Edinburgh EH1 1HN, Scotland, UK
    Tel: +44 131 650 2785
    Fax: +44 131 650 2784
    Email: eddie@ccir.ed.ac.uk

    Fourth Annual Summer School to be Held in Budapest

    Following on from last year's summer school on multilinguality held in Edinburgh, the Fourth European Summer School on Language and Speech Communication (better known as the ELSNET Summer School) will be in Budapest from July 8-19, 1996. The topic will be Dialogue Systems.

    There has been a growing interest over recent years in the theoretical and practical issues associated with the design and use of computer systems which are able to participate in spoken or written language dialogues. Some of the questions which have occupied researchers include the following:

    These and many other questions relating to dialogue systems will be addressed in the Summer School in courses which will be taught by leading practitioners in the field. Courses will include the following topics: speech input and output in dialogue; empirical foundations for dialogue design; multimodal systems; natural language input and output in dialogue; dialogue modelling; human factors in dialogue design; architec-tures and systems design; historical overview; types of dialogue; prosody in spoken dialogue; evalua-tion of dialogue systems; and practical applications.

    The courses will be a mixture of short plenary sessions on particularly diffi-cult or contro-versial topics or providing surveys, and weekly courses of 5x2 hours or 5x1 hours held in plenary or in parallel. Several of the courses will set practical exercises, and there will be ample opportunity for students to present their own work. As is fitting in a Summer School on dialogue, participants will be encouraged to play an active part in the learning process. Background knowledge in a relevant area such as linguistics, speech processing, artificial intelligence, com-puter science or psychology would be useful, but no prior experience in the area of dialogue systems will be assumed.

    The Summer School is open to under-graduate students, PhD students, postdocs, and staff members from academic and industrial sites. Fees for students will be 100 ECU; for academic scholars, 200 ECU; and for industrial participants, 400 ECU. A limited number of grants can be made available for participants from Central and Eastern Europe.

    Members of the programme committee are: Niels Ole Bernsen (Roskilde University, Denmark), Norman Fraser (Vocalis, UK), and Klara Vicsi (Technical University of Budapest, Hungary). The main sponsors and supporters are the European Network in Language and Speech (ELSNET), the Copernicus-LRE project ELSNET goes East, the European Speech Communication Association (ESCA) and the European Chapter of the Association for Computational Linguistics (EACL). Local support is provided by SUN Europe and the Technical University of Budapest.

    Characteristic of the school is its small scale (maximum of 90 participants), and its selection of a topic which has interest and relevance for both the speech and the computational linguistics communities.

    Formal registration will be open from January 15, 1996, and registration forms (and more information concerning the programme) will be distributed by that time. Registration deadline will be around 15 April - 1 May, 1996. Expressions of interest and requests for further information may be sent to:

    Klara Vicsi
    Email: school@khmk.bme.hu

    The Text Encoding Initiative

    Lou Burnard, Oxford University Computing Services

    The Text Encoding Initiative (TEI) is an international cooperative research effort, the goal of which is to define a set of generic Guidelines for the representation of textual materials in electronic form. The project is concerned not only with what textual features should be encoded (i.e., made explicit), but also how that encoding should be represented for loss-free, platform-independent, information interchange. By providing a description of information which is independent of realisation or media, the TEI scheme enormously facilitates the construction and exploitation of multimedia technology.

    The prime deliverable of the project is more than 400 textual feature definitions, expressed as SGML (Standard Generalized Markup Language) elements and attributes, with associated documentation and examples. These elements are grouped into tag sets --- a tag set being a collection of definitions --- and together constitute a modular scheme which can be configured to provide hardware-, software-, and application-independent support for the encoding of all kinds of text in all languages and of all times.

    As an SGML application, the TEI scheme requires the existence of some kind of document type definition (DTD). Several DTDs are defined, and each of these may be tailored to the needs of a particular application, by the combination of different tag sets, giving the user the opportunity of building a DTD which matches his or her requirements, while at the same time constraining the possibilities in a way that facilitates interchange.

    The TEI tag sets are based on, but not limited by, existing encoding practices; they are designed to be both comprehensive and extensible. They are collectively documented in a substantial reference manual, the Guidelines for Text Encoding for Interchange, which appeared in May 1994. This 1,400- page manual is available both in paper and electronic hypertext form as well as over the Internet, in a variety of formats.

    Because of the enormous size of the document, the TEI plans to make available a number of smaller introductory tutorials focused on particular applications areas. Two such have already appeared: one dealing with terminological systems, and the other on encoding of manuscript transcriptions. A third tutorial, known as TEI Lite, documenting a subset of some 200 elements from the TEI scheme, has already been used in two electronic publishing projects and is in use at electronic text repositories at the Universities of Oxford, Virginia and Michigan, and elsewhere. The TEI Lite tutorial is available on-line.

    While the EC-funded EAGLES project is committted to the TEI scheme, the real proof of its effectiveness will come only with its wide-spread adoption, tailored to the need of individual projects. As far as can be judged from the long list of early implementors, such evidence will soon be forthcoming.


    This note is extracted from a fuller description of the TEI scheme. Additional information, including the TEI Lite tutorial, is available here.

    A browsable version of the Guidelines for Text Encoding for Interchangeis also available.

    New book on the Text Encoding Initiative

    Kluwer has recently published a new book entitled Text Encoding Initiative: Background and Context, edited by Nancy Ide and Jean Veronis.

    Table of contents:

    Charles Goldfarb: Preface
    Nancy Ide and Jean Veronis: Introduction

    Part I: General Topics

    Nancy Ide and C.M. Sperberg-McQueen: The Text Encoding Initiative: Its history, goals, and future development
    C.M. Sperberg-McQueen and Lou Burnard: The design of the TEI encoding scheme
    Lou Burnard: What is SGML and how does it help?

    Part II: Document-wide encoding issues

    Harry Gaylord: Character representation
    Richard Giordano: The TEI header and the documentation of electronic texts
    Dominic Dunlop: Practical considerations in the use of TEI headers in large corpora

    Part III: Encoding specific text types

    David Chisholm and David Robey: Encoding verse texts
    John Lavagnino and Elli Mylonas: The show must go on: Problems of tagging performance texts
    Robin Cover and Peter Robinson: Textual criticism
    Daniel Greenstein and Lou Burnard: Speaking with one voice: Encoding standards and the prospects for an integrated approach to computing in history
    Stig Johansson: The Encoding of spoken texts
    Alan Melby: E-TIF: An Electronic Terminology Interchange Format
    Nancy Ide and Jean Veronis: Encoding dictionaries

    Part IV: Special encoding mechanisms

    Steven J. DeRose and David Durand: The TEI hypertext guidelines
    D. Terence Langendoen and Gary F. Simons: Rationale for the TEI recommendations for feature-structure markup
    David Barnard, Lou Burnard, Jean-Pierre Gaspart, Lynne A. Price, C.M. Sperberg-McQueen, Giovanni Battista Varile: Hierarchical encoding of text: Technical problems and SGML solutions
    SGML/TEI Bibliography by Robin C. Cover


    Hardback ISBN 0-7923-3689-5; Paperback ISBN 0-7923-3704-2, Number of pages: iv + 242 pages (includes an SGML/TEI bibliography).

    Prices for hardback: US$ 99.00; 145.00 Dutch guilders; 60.00 pounds sterling.

    Prices for paperback: US$ 45.00; 75.00 Dutch guilders; 29.00 pounds sterling.

    Attention All Software Developers!

    LISA, the Localisation Industry Standards Association, has recently announced the availability of the its Standardised Quality Assurance (QA) Model for Software Localisation. The QA Model provides guidelines and metrics for localisation quality in order to assist with the localisation research, production, and quality control processes for software developers and service vendors. The documentation and templates help clients and vendors implement a synchronised QA system at multiple production sites. The system includes features such as localisation process quality control; QA checkpoints; printed and on-line documentation and computer-based tutorials; and document quality checklists for language, formatting, and functional art and design. The cost of the system is $275 (or $175 for LISA members).

    Also available on CD-ROM is the LISA Showcase, which is a multimedia reference database containing detailed information about products, processes, standards, and methodologies for the localisation and internationalisation business. The LISA Showcase is a directory and catalogue of localisation and language processing tools, supplier profiles, products and guidelines. It also contains industry trade journals, independent product reviews and European Community localisation and software development project reports. The database will run under Windows on standard hardware, and has a state-of-the-art user interface that permits cross-referencing of data for easy information access. Published twice per year, the LISA Showcase will provide precise and up-to-date information, making it the most comprehensive guide to localisation and language processing tools available today. The cost of one CD is $250. (LISA members and subsidiaries of LISA member companies receive substantial discounts.)

    For details about the LISA Quality Assurance Model, the LISA Showcase, or LISA membership, please contact:

    2 bis rue Ad.-Fontanel
    CH-1227 Carouge
    Geneva, Switzerland
    Fax: +41 22 301 5761
    Email: manobile@divsun.unige.ch

    Future Events

    March 7-9, 1996: Les langues et leur images. Neuchatel, Switzerland. For information, contact IRDP, Fbg de l'Hopital 43, CH-2007 Neuchatel, Tel: +41 38 24 41 91, Fax: +41 38 25 99 47.

    April 11-12, 1996: Second ACM/SIGCAPH Conference on Assistive Technologies. Vancouver, Canada. Deadline for submission of papers: Oct. 17, 1995. For information, contact: David Jaffe, Dept. of Veteran Affairs Medical Center, 3801 Miranda Avenue, Mail Stop 153, Palo Alto, CA 94304, Email: jaffe@roses.stanford.edu.

    April 24-26, 1996: Workshop on Speech and Image Understanding (3rd Slovenian-German and 2nd SDRV Workshop). Ljubljana, Slovenia. Deadline for paper submissions: Feb. 15, 1996. For information, contact: Nikola Pavesic, Univ. of Ljubljana, Faculty of Electrical & Computer Engineering, Trzaska 25, SI-61000 Ljubljana, Slovenia, Email: nikolap@fer.uni-lj.si.

    June 4-6, 1996: International Conference on Natural Language Processing and Industrial Applications. Moncton, New Burnswick, Canada. For information, contact: Dr. Chadia Moghrabi, GRETAL, Departement d'informatique, Université de Moncton, Moncton, NB, Canada E1A 3E9, Email: nlp-ia@umoncton.ca.

    June 28, 1996: Second meeting of the Special Interest Group in Computational Phonology (SIGPHON 96). Santa Cruz, CA, USA. For information, contact: Richard Sproat, AT&T Bell Laboratories, Room 2d-451, 600 Mountain Avenue, Murray Hill, NJ 07974, USA, Email: sigphon@reseach.att.com.

    July 8-19, 1996: The Fourth European Summer School on Language and Speech Communication. Budapest, Hungary. For information, contact: Klara Vicsi, Email: school@khmk.bme.hu.

    July 15-19, 1996: The Auditory Basis of Speech Perception, (ESCA Tutorial and Reseach Workshop). Keele University, UK. Deadline for submission of abstracts: Nov. 15, 1995. For information, contact: ESCA Workshop, Dept. of Communication and Neuroscience, Keele University, Keele, Staffordshire ST5 5BG, UK, Tel/Fax: +44 1782 583055, Email: cob03@keele.ac.uk.

    August 4, 1996: Fourth Workshop on Very Large Corpora (WVLC-4). Copenhagen, Denmark. Deadline for paper submissions: Apr 10, 1996. For information, contact: Eva Ejerhed, Dept. of Linguistics, DGL, Univ. of Umea, S-90187 Umea, Sweden, Email: WVLC-4@ling.umu.se.

    August 5-9, 1996: International Conference on Computational Linguistics (COLING-96). Copenhagen, Denmark. Deadline for paper submissions: Dec. 15, 1995. For information, contact: Bente Maegaard, Email: bente@cst.ku.dk.

    August 12-16, 1996: 12th European Conference on Artificial Intelligence (ECAI-96). Budapest, Hungary. Deadline for workshop proposal submissions only: Nov. 1, 1995. For information on ECAI workshops, contact: Elisabeth Andre, DFKI, Stuhlsatzenhausweg 3, D-66123 Saarbrücken, Germany, Email: ecai-96-ws@dfki.uni-sb.de, or look here.