This concluding section takes the form of a commented summary of the contents of the report. Each section will be summarized in turn, and an attempt made to pick out the main points which emerged from that section. Almost inevitably, those points will often touch on what the participants saw as problem areas. Repeating them here should not be misconstrued as hostile criticism.
We noticed that the SdT was responsible for dealing with an enormous volume of text, producing more than a million pages of translation a year. We also noticed, both here and elsewhere, that the documents sent for translation are very frequently the product of many different authors, and their authors are often not writing in their mother tongue. Consequently, translators are called upon to a degree unknown in other translation services to deal with language which is not of native speaker quality.
We noticed too that because of the equal status of Union languages, the term "translation" was only meaningful as long as a document remained inside the Commission. Once it left the Commission there was no indication of what the original language of a document might have been, and all language versions enjoyed equal status.
The SdT is a very large and very complex organisation, with around 1,700 staff organised into departments according to requesting services, into Units according to language, and into horizontal groupings providing a variety of support functions. In addition to the in-house staff, heavy use is made of freelance translation. If we also take into consideration the relations with the requesting services, it is hardly surprising that two topics, information flow and work flow, surfaced very frequently throughout the study, with work flow, both inside and outside the SdT, in particular, often being perceived as a major source of difficulties.
The SdT is heavily computerized, and many computer based tools are already available as aids to management and as aids to translation. New projects are continually being defined and developed, and consideration of future plans took us into some of the problem areas for which solutions are sought.
Many of these problems are the almost inevitable result of an organisation which is continually growing in size, and which is also having to face the challenge of moving from paper-based information processing to electronic treatment. Thus, the major problem of document identification (the inter-relationships between the SdT internal document identification code, the COMM, SECC and C document numbering systems) can be seen both as the result of historical process and as a problem calling for urgent resolution in the light of new possibilities offered by electronic archiving.
In term of solutions to particular problem areas, it was particularly noticeable that a pragmatic attitude had already led to ingenious and innovative solutions. As examples, we might cite the use of AVIMA to generate multilingual versions of calls for tenders, thus relieving the SdT of the necessity to produce a second million pages of translated text every year, the development of SEI-BUD to deal with the production of the budget in the form of a single integrated document which could be worked on simultaneously by all of the three main actors in its production, and the use of TMan to replace repetitive text in the monthly Bulletin of Activities, cutting translation needs by about half.
On a more general level, we noticed the growing use of the SYSTRAN machine translation system by officials outside the SdT to obtain rough translations without passing through the translation services, and the use of SYSTRAN within the SdT to get rough translation of background documents or, on a more limited scale, as a starting point for more polished translation, as well as a means of obtaining terminology or CELEX references.
Amongst the more recent products being evaluated for integration into the SdT's array of translators' aids tools, translators' workbench systems and local terminology management systems were prominent. A lively interest in voice dictation systems and the possibilities they might offer was very widespread, both amongst the translating staff and amongst the staff from the horizontal units.
On the organisational side, we noticed that a number of schemes for Inter-Institutional collaboration were under-way or being planned, ranging from quite modest attempts to collaborate on training and recruitment to more ambitious projects such as a single terminology base for the European Union, or to resolve the document identification problem.
Starting once more from the original documents sent for translation, we noticed that the demand is not only large, it is also very variable, and that, large as it is, it does not reflect the total amount of translation done within the Commission. As well as the use of SYSTRAN already mentioned, it is not unusual for officials to translate themselves, or to arrange for translation to be done without going through the SdT. We also saw that this can lead to problems, since the SdT is perceived by the outside world as being responsible for the quality of all translation, without distinction, and, in any case, the outside world normally has no way of knowing whether a text was produced by the SdT, by the requesting service or by SYSTRAN. Translators sometimes understandably resent poor translation appearing in official documents.
Most documents are still transmitted on paper. Strenuous efforts are in hand to increase the number of electronic submissions, and all translations are returned in electronic form, no matter how they have been submitted. Using e-mail for document transmission is not yet part of the normal way of life in the requesting services, although it is gradually becoming more familiar, and problems like a requester not knowing how to print out his urgent translation are becoming rarer. It should be noted though that a considerable proportion of documents originate from outside the Commission itself, from the Member States or elsewhere. In these cases, whether or not a document is transmitted in electronic form may not be within the control of the requesting service.
The requesting services do not always use the same text-processing system as does the SdT. Conversion programs exist, and the use of a standard for document preparation, EUROLOOK, is being promulgated not only in order to ensure a uniform appearance of Commission documents but also in order to guarantee convertibility.
Apart from these banal, but nonetheless important questions, a number of issues surfaced here that have to do with engineering the work flow between the SdT and the requesters, and, to some extent, within the SdT itself. Many of these have to do with two problems which were raised over and over again, the problem of version control and the problem of acquiring all appropriate reference material for a translation. Others, perhaps even more intractable, had to do with the difficulties of planning, especially in certain departments where the workload is extremely unpredictable and urgent work coming in can have serious consequences on planning that has already been done. This issue is complicated by the fact that there is no central policy on how to resolve conflicts. Yet other work flow problems have to do with at what point in the document production chain certain operations on a text might best intervene. For example, if there are problems with the source text, can they be trapped already when the document arrives in the department and resolved before the document is passed on down the chain to the translation unit or even the individual translator? Could some work on terminology be done high in the chain, and thus avoid reduplication of work at the individual translator level? Where should any preprocessing be done? Some of these questions also have a bearing on the issue of relations with the requesting services, who will obviously prefer to be asked a question once than to be asked the same question by a number of different translators who are all working on the same document.
Not surprisingly, information flow and communications issues also appeared largely in discussing the work of the Department Heads, as did the use of management tools like SUIVI. The question of encouraging standardization of documents also came up again here. Apart from the need to be in constant communication with the requesting services, the most commonly cited difficulty was that of trying to ensure that translation into the different languages proceeded in parallel. A new issue that came up in discussing the work of the Department Heads was the possibility of distinguishing different kinds of translation for different purposes: of suggesting to a requester, for example, that a quick oral summary might suffice for his needs, or that he might consider asking for a SYSTRAN translation. One striking aspect of the work of the Department Heads is the rich variety of activities and responsibilities they undertake in addition to direct task of liaising with the requesters and seeing that the translations get done.
Quite a large proportion (around 25) of translation work is dealt with by freelance translation. A number of participants pointed out that it was difficult to determine the economics of using freelance translation, since direct comparison of costs makes little sense, but that, on the other hand, since it is simply not feasible to maintain a full-time permanent staff capable of dealing with the peak load, use of freelance work was inevitable. Freelance work is also used to deal with specialized texts, and we noticed that there is a long term policy to identify families of documents that could systematically be sent for freelance translation, thus easing the general work-load. When the interviews on which this study is based were beginning, a change in the procedure for the selection of freelance translators had recently come into effect, and a number of those interviewed expressed worries about how the change would affect the quality of freelance work. Given the very large numbers of people involved, of language combinations and of subject areas, no really thorough testing of those who had proposed themselves had been possible, and freelance translators were in effect being field tested by being used. It will take some time before a well founded judgment of the effects in the change of procedure can be formulated.
Use of freelance translators once again raises issues of communication and of work-flow, as well as specific issues of access to the various facilities and support services available to the in-house translator.
Another way of easing the burden on the translation services is to make use of machine translation. We noticed that the Commission has had a long standing interest in this area, and that the SYSTRAN machine translation system is now quite widely (and increasingly) used by officials outside the Translation Service. The SdT remains however the largest single user of SYSTRAN, and translators use it to obtain rough translation of background documents, to obtain CELEX material and other reference material and as a way of accessing terminology as well as sometimes to obtain a raw translation as a basis for more polished work.
A document which follows the conventional track through the SdT will go from the Department planning service to the Translation Units, which are organised around the individual languages. The Unit Heads, whose job is to make sure that translation of appropriate quality into their language gets done on time, either within the Unit or through freelances, emphasized the need for local management tools. They also stressed the utility of standardization in document preparation. Since the Unit is responsible for the quality of all translation done within the Unit, it is perhaps not surprising that several of the Unit Heads raised the question of quality control of freelance work. Quite naturally, the problem was felt most acutely in those Units where there was not the capacity to do much revision of freelance work. The Unit Heads carry quite a lot of administrative responsibility in addition to work directly related to seeing to it that translations are produced. Their vulnerability to interruptions frequently meant that they limited their own translation work to short translations or to very urgent work, although all seemed to be heavily involved in revision.
The document will find its way, eventually, to the desk of the individual translator, who may have been allotted that particular document or who may have chosen it out of the work waiting to be done, depending on the policy in his particular unit. His first task is one of research: of finding the reference materials, the terminology and the citations needed to complete this particular translation. Typically he will make heavy use of EURODICAUTOM and/or of CELEX at this point. These two aids to translation are used far more heavily than any others. EURODICAUTOM, in particular, was frequently cited as the absolutely essential tool. This is not to imply that both translators and terminologists were unaware of the weaknesses of EURODICAUTOM, many of which are inevitable consequences of its age and of its size. Creating a replacement for EURODICAUTOM or even simply bringing it up to date and cleaning it up is a mammoth task which also presents the challenge of not forfeiting the immense riches already contained therein.
A potentially rich source of future terminology is the local terminology created either by the individual translator or by groups of translators working on a common project or in the same unit. The PC-based client-server architecture of the computing infrastructure offers the possibility of using local terminology management systems in order to exploit the local reserves. However, this brings with it questions of validation and of how to encourage the individual translator to make his terminology available which are currently being attacked.
There are almost as many ways of tackling the process of translation as such as there individual translators. Most still dictate their translations, although substantial numbers now work directly onto a key-board. Whether dictating or typing, some will prepare a very rough first draft and go back over it a number of times, others will produce an almost final version straight away. Some will work right through to the end before carrying out any modifications, others will continually back-track over what has already been done. Both preferred ways of working and the urgency of a translation influence the choice between dictating and typing. Both those who preferred typing and those who preferred dictating showed a lively interest in voice dictation systems, which seemed to many to offer the best of both worlds, the speed and ease of dictation being combined with immediate availability of a text in electronic form. Whether dictating or typing, it is important to be able to concentrate and not to suffer from distractions which also come in many forms: ergonomically unsatisfactory working conditions were amongst those frequently cited.
Almost inevitably, communication and work-flow problems were also present in the life of the individual translator. Where there was a problem with the original text it was sometimes difficult to find the person in the requesting service who could resolve it, and this could have consequences on the flow of work. Translators also sometimes found it frustrating that each individual translator was trying to resolve a problem which they felt could have been dealt with centrally, either in the requesting service before the document was transmitted or higher up in the chain inside the SdT. Translators were also aware of the risk of work being done more than once, when a document contained recycled text or when version control had broken down. In this context they were keenly aware of the possibilities offered by electronic archiving, and eager for the development of good indexing and search tools.
Translators may be involved in a very wide range of activities outside the work of translation, ranging from involvement in the Permanent Delegation of Translators, which plays a highly visible role in representing staff views to the management, to acting as computing, terminology or training correspondent or helping with the evaluation or development of new tools. Although such activities are appreciated as adding interest to the working day, those who are involved in them sometimes feel that they are resented by their colleagues as not pulling their full weight in the production line, or worry that their involvement is not directly reflected in any quantified statement of their own productivity, such as the number of pages of translation produced.
The next critical link in the document production chain is the secretaries, who type any text which has been dictated and are responsible for modifications in subsequent versions or as a result of revision. The SdT experiences some difficulty in keeping secretarial staff, partly because the SdT is the most typical first post for new arrivals, partly because audio and copy typing is not very interesting work. Some units consequently have recourse to freelance typists, which can once again lead to work-flow problems, especially in those cases where the freelance typists return all but the final versions of their work on paper rather than on diskettes, so that corrections and modifications cannot be done within the Service but have to be sent back to the freelance. This problem is exacerbated by the fact that the original author is not usually easily available to the freelance typist to clear up doubts or questions.
Policy about revision varies enormously from unit to unit. There are units where every document is revised: other units are obliged to be selective. In some units, the revisor is always a more senior person, in others revision is a co-operative effort between equals based on four eyes seeing more than two. Freelance work can be a special worry, especially in those units where it is only possible to carry out spot checks. It was noticeable that revisers, perhaps even more than translators, need good tools and resources for documentation and research.
A finished translation is returned by e-mail to the requesting service. Typically, the SdT will receive no feedback on whether the job has been well done. The requester himself obviously cannot comment on versions into languages he does not know, and there is no obvious way of obtaining feedback from the end users of the translation. Many of the participants found this frustrating, and some had found ways of their own to bridge the gap, for example by occasionally attending meetings of the Committees who used the documents. Even so, there was a very common perception of a great void at both ends of the document production chain, to which the many suggestions that have been made both officially and unofficially of ways to bring requesters and translators closer together bear witness.
Terminology is obviously of critical importance, and the heavy use of EURODICAUTOM has already been mentioned. The difficulties of maintaining and updating EURODICAUTOM were of particular concern to the terminologists, although it was pointed out that the recent implementation of batch access to EURODICAUTOM is of some help here. There was also considerable discussion of what a successor to EURODICAUTOM should look like. The issue of how best to validate and exploit local terminology naturally surfaces again in this context. The terminologists also make heavy use of other data bases, with RAPID, the data base of press releases, and various corpora of newspapers being frequently mentioned. Both terminologists and some individual units have frequent contact with specialist bodies outside the Commission, in some cases in the form of systematic collaboration on terminology acquisition and validation. Some terminology work is contracted out to freelances, essentially as piece-work. This can lead to problems of quality control. Individual translators can request help with terminology through the terminology help desk, whose staff point out that experience with EURODICAUTOM teaches them "tricks" of access, for example through the subject codes, which can considerably enhance the hit rate and reduce noise.
The services of specialist documentalists are also available, and we noticed that the role of documentation specialists grows ever greater with the explosion of the information society. The documentalists make very heavy use of data bases, both internal to the Commission and on the exterior. One problem with access to data bases is that each has its own interface and mode of access: the aid of specialists here was much appreciated. For acquisition purposes, the documentalists are in permanent contact both with the requesting services and with national bodies. The clientèle for specialist documentation help is very wide: translators and terminologists, officials outside the SdT, interpretors and freelances all call on these services. This can lead to problems of knowing what requests should be fulfilled, and even of knowing who has a right to what documentation.
Training is a constant requirement within the SdT, but we noticed that the migration to a new computing architecture had required a special effort. Also, the introduction of any new tool brings with it training needs, even if the tool is only being temporarily installed for evaluation purposes. In addition to special efforts triggered by change, regular courses are offered on the use of CELEX, on project management and on the new tools already in service and their use.
Help and advice on computing facilities and on the use of tools is available through the computing correspondents and the computing help desk. There are both full-time computing correspondents who are also help desk staff and part time computing correspondents in the departments and in the individual units.
Staff from both the horizontal support units and from the translation units co-operate in the evaluation of new tools, amongst which translators workbenches and local terminology management systems are currently receiving attention. A number of issues are related to the evaluation and integration of new tools. First, there is some danger of unrealistic expectations from the side of the users. The unwary might think that a local terminology management system actually contained some terminology, rather than being a black box into which the terminology must be put, or that a translation memory came equipped with a set of example translations. Secondly, the market as a whole is geared much more to the individual translator or to a small group of translators than to the special needs of a large translation service. Thus, issues like validation of terminology or of the translations to be stored in a translation memory are largely neglected by the products on the market. Furthermore, integration of a new product into an already existing large computing service causes special problems. The individual manufacturer is, at best, concerned with compatibility only across his own range of products. A working translator cannot be expected to familiarize himself with a whole variety of different products, each of which has its own interface, its own set of commands and its own peculiarities. The EURAMIS interface created within SdT and currently on experimental release attempts to palliate these problems by providing a single uniform interface. A project of the same name jointly directed by D.G. XIII and the SdT aims at giving access to a wider range of tools and resources through a single uniform interface.
The evaluation of new tools will now be further pursued in the translation workshop (Atelier de Traduction) in Brussels and the Modernization Network (Réseau de Modernisation) in Luxembourg. These newly created organisational structures will put the new tools to use in a productive work environment.
The SdT has its own computing service. Many of the concerns of this service have already been touched upon. It is however worth adding here that market forces also have an impact on the effort to keep all languages equal. It is, for example, easy to find a text-processor with an English or French interface, but much more difficult to find one with a Greek or Dutch interface.
The size and complexity of the computing infrastructure means that the SdT has to adopt a somewhat conservative policy towards updating the services offered. Quite apart from the capital cost, the translator using the services cannot be expected to learn a new system every few months, and his productivity would suffer should he attempt to do so. Relative conservatism combined with the new PC-based client-server architecture is responsible for some tensions between the end users and the computing service. Even though the vast majority of the potential end user population have no ambition to become computing experts, a lively and articulate subset are increasingly aware of the possibilities offered by a PC. The computing services must struggle to find a compromise between giving the knowledgeable what they want, giving the majority what they need and are prepared to use, maintaining compatibility of both computing resources and linguistic resources and being able to ensure maintenance of what is offered: a delicate and challenging task.