The translator is the critical actor in the whole multilingual document production chain. Translations to be done are distributed amongst the in-house translators in two main ways.
In some groups, the Head of Unit will check who has dealt with previous versions of the document, who has worked on similar texts or who has spare capacity and will allocate the translations accordingly. In others, the translations are distributed on a self-service system, with the translator himself deciding which document to take. Which system is preferred seems to be as much a matter of cultural preference as anything else.
A recent innovation has been the organisation of workshops within the SdT to reflect on the translation process and how it works. These workshops distinguish three states as part of the translation process:
We shall follow this same scheme in organizing the rest of this section.
Research essentially involves finding references and reference material and solving translation problems, including the problem of terminology.
A properly prepared translation request is accompanied by all the relevant reference material. However, the translator may discover once he starts work on a document that some reference material is missing. There may also be hidden references: quotations, for example, where the source is not given. His resources for finding missing reference material include the archives of previous documents and their translations, the various documentary and bibliographic data bases, the library and the services of a documentalist and the requesters themselves.
Many translators will search the relevant data bases themselves. Others will ask a documentalist to do it for them. One data base very frequently used is the CELEX data base of legislation, court judgements and preparatory documents. Two problems were mentioned with this data base. The first concerns the data itself. Some of the older entries from before 1989 are encoded in an impoverished character set. A great deal of wordprocessing work is needed to recover the text in its original form. The second concerns the interface available on the older Q-One systems, which is felt to be very unuserfriendly.
It should also be pointed out that, inevitably, CELEX is never up-to-date: there is a delay of up to three months before the contents of the Official Journal find their way into CELEX, and of about eighteen months for court cases. There is a difference too between when the orginal may become available and when translations of it are available. Thus, although a recent decision of the Court of Justice may result in court rulings becoming available more quickly, there may well still be a substantial delay in the translations becoming available.
One way of procuring a translation of the references to previous documents included in a text is to ask for a SYSTRAN translation: an automatic search of CELEX during the machine translation process results in a list of references appended to the SYSTRAN raw translation. Many translators mentioned using SYSTRAN for this purpose alone. This use of Systran will be replaced by the EURAMIS integrated interface currently under test in the SdT, which allows CELEX search alone. Although it is comparatively rare for translators to use SYSTRAN as a starting point for the translation itself, it is also occasionally used to get a rough translation of background documents.
Terminology research also makes up a major part of the translator's work.
A primary source of terminology is EURODICAUTOM, the Commission electronic terminology base. The base can be consulted through a single query or by sending a list of queries. Recently it has become possible to submit a text to SYSTRAN and receive in return a list of terminology extracted from the text, based on the SYSTRAN dictionaries enriched by the EURODICAUTOM data.
However EURODICAUTOM is a long-established data base, and using it reveals some problems. The two most frequently mentioned were the amount of noise in the form of numerous responses to a query, and the problem of knowing what the original language of an entry had been. This latter perhaps merits a little explanation: when a new language is added to the entry, the terminologist will normally be working from one of the languages already there. There is often no indication in the entry which language was the original, although sometimes the reference field may give details which will in effect indicate the original language. But this is far from being the usual case, and so it is possible for a Portuguese equivalent, say, to be constructed on the basis of a French equivalent, which was itself constructed on the basis of an English term. If terminology were such that chaining from one language to another always produced the same result, this would not matter. But unfortunately this is not the case, and the end of the chain may turn out not to be the correct terminological equivalent to the term at the beginning of it. In fact, this is a simplified statement of the problem, which may go further: in some cases, there may not be one single original language, since several entries in different languages may have been taken as the basis for different new entries.
Despite these problems, EURODICAUTOM is very heavily used: almost everybody mentioned it as very high up on the list of tools they used, although I did meet just one person who claimed never to have consulted either EURODICAUTOM or CELEX.
The terminologists and those responsible for EURODICAUTOM point out the same problems, and add the difficulty of maintaining and updating a term bank of this size and age. All the questions of validation which have already been mentioned re-appear in this context, and special problems appear when the possibility of cleaning-up and consolidating the entries already in the base is considered. When EURODICAUTOM gives multiple responses to a query, it may be that some or all of them can appropriately be synthesised into a single entry. This is difficult and demanding work, which takes time. The resources needed are substantial, and are not always available. A recently installed facility which allows consultation of EURODICAUTOM in batch mode with a list of terms may to some extent facilitate the task of maintenance.
The Council terminology bank TIS is also available for consultation and is used. Daily up-dates to it are received.
Many translators create their own terminology data base, either manually on paper or through PC tools. Some units share their terminology across the unit and some groups encourage the creation of requester-oriented terminology. Almost all individual translators consult their colleagues, at least from time to time. This leads to the creation of local terminology resources within individual translation units or departments. We have already mentioned the potential value of such local terminology as material to be included in the central resources, but have also mentioned some of the problems of sharing and of validation.
Some units have established cooperation with specialist organisations in the Member States. They consult them on terminological questions and obtain feedback from them on the translations done.
Some translators make a practice of going at least occasionally to meetings of the Committees which use their documents. In this way they get to know the national experts and can contact them for terminology problems, and can also get feedback on their translations. One suggestion made was that lists of those who had attended the meetings might be systematically sent to the translation units so that they could solicit feedback if they wished. One translator mentioned that seminars organised with the national experts were of great help.
When all else fails (or even, according to the people responsible for running it, before), the translator may appeal for help to the terminology helpdesk, run by AGL 3. The terminologists on the helpdesk point out that there are clever ways to search EURODICAUTOM which help to increase the success rate and to cut down noise. Some of these involve use of the subject codes, which, although necessarily intuitive and therefore open to interpretation, one can become expert in. They suspect that translators become discouraged if they do not find the information they need through simple means, and are also sometimes embarrassed to admit that they do not know any of the tricks. They suggested that it might be helpful to organise short training sessions on clever ways to search EURODICAUTOM.
Ways of working when actually translating vary greatly.
At the moment, most translators work alone. But in some groups, translators are encouraged to work in cross-language teams, at least for some documents. Where this happens, translators find it helpful to be able to look at the solutions that their colleagues have found to particular translation problems.
Many translators dictate. But what they dictate varies: some dictate a very preliminary first draft which they then tidy up and revise on paper, others dictate an almost finished translation which only needs subsequent checking.
Others prefer to work on a screen, still others sometimes dictate, sometimes work on a screen, depending on the type of document or the urgency of the translation.
The motivations for using one method or the other are almost as many as there are translators. There are those who are convinced that dictation leads to better translation, either because the musicality of the language plays a greater role than when one types or because dictating leads to a simpler and clearer style, with shorter, more common, words and less complex sentence structure. Others dictate because their typing ability is poor, or because dictating is faster than typing. Even those who prefer to use a keyboard show a tendency to dictate when a piece of work is particularly urgent. Amongst those who prefer to use a keyboard, there are those who are technically minded and like using machines, those who always found dictation a painful and inefficient process because they frequently want to go back over passages that have already been translated, and those that were simply trained to use keyboards and find it hard to imagine doing anything else.
It was pointed out that choosing to use a PC rather than dictating can cause the translator to become involved in problems that otherwise fall onto the secretarial staff. He may, for example, spend time persuading a recalcitrant printer to behave as it should, or figuring out how to create and manipulate tables. There is also a connection here to the page counting issue mentioned earlier. If a translator types his own translation, he is saving secretarial work for his unit, but this is not directly reflected in his own productivity. Many translators are quite slow typists, and will therefore prefer to dictate in order to increase the number of pages produced, even if otherwise they might prefer to work on a screen.
Whatever the method chosen, it is important to be able to sustain a certain level of concentration. (Those who have been involved in the translation workshops talk of a "translation state" which takes time to enter and which is difficult to recapture once it is broken). Thus, interruptions are unwelcome, and distractions infuriating.
Distractions can come in many forms. Several people talked of the distraction caused by being in physically and ergonomically unsatisfactory environments: screens placed by windows, chairs and tables at the wrong height, construction work going on directly outside and so on. Others talked of the importance of what is on the screen: a flashing icon or a mail system pinging when a message arrives can break concentration and interrupt the flow.
For those who work on screens and perhaps use several tools at the same time the quality of the interface is also important. Interfaces must be integrated, and the same manipulation of the keyboard or mouse must always mean the same thing, no matter what tool is being used. This poses problems when commercial products have to be integrated into the computing environment. Computing environments must also be relatively stable. A translator cannot be expected to accustom himself to a new environment every few months. Furthermore, the capital investment involved in changing the computing environment is substantial. The SdT is heavily computerized, with more than twenty minicomputers, 1,200 PC's (which will become 1,500 by the end of 1995) and all the peripheral equipment such a set-up implies. Upgrading the environment also means adaptation of existing software. For all these reasons the computing environment in the SdT tends to change slightly more slowly than the state of the art. Thus, for example, although WordPerfect 6 is now available and is in many ways a superior Windows product to WordPerfect 5.2, migration to WordPerfect 6, or, possibly, some other word processor, will not happen until the end of 1995 or even later. Similar considerations determined the choice of DOS 5 rather than DOS 6.
The tools used by the translator during the process of translation are, as the above implies, those associated with document creation: the dictaphone or a text-processing system being the main ones. Many translators use spelling checkers to check their finished text. None use grammar checkers to check translations.
There was lively interest in voice dictation systems both amongst those who dictate and amongst those who type. These systems seemed to many translators to combine the best of both worlds. The translator would be freed from the pain of typing in his text, but nonetheless would have it available immediately on the screen, so that he could see what he had said, could scroll back to earlier in the text and so on. Many of those interviewed thought that the current limitations of voice dictation systems would not be a great drawback to their use in a translation service. That they are only capable of recognizing discrete speech is less likely to perturb people who are used to dictating, and that the system has to be trained for each voice that it has to recognize is not a great hindrance in a set-up where most people have a machine of their own. One group of translators were, though, very firmly opposed to the idea of giving voice dictation systems to the authors of the original documents.
It happens fairly frequently that a translator will find, whilst working on a text, that there are passages of the original which are ambiguous or which he does not understand. In this case, he needs to contact the author of the document to get clarification. This can prove problematic. The request slip asking for the translation contains the name of a contact person, but this will often be a secretary in the requesting service rather than the author. Once the author has been found, it can prove difficult to explain the nature of the problem to him. A common reaction that was quoted several times was for the author to say that the reader would understand anyway, without realising that the translator cannot translate what he himself does not understand.
An anecdote may illustrate this point and also provide some light relief. A recruitment test, drafted in French, contained the question ``Qu'est-ce que c'est le GATT?'' The multiple choice answer, amongst wrong answers, of course contained the correct ``L'accord général sur les tarifs douaniers et le commerce''. When the test was sent for translation, its author was astonished to be told that the question could not be used, since, once it was translated into English, the question contained its own answer and therefore favoured English speakers.
The most common way round this is to phrase the question as ``So what you really mean is ...'' and get an answer that way. Once again, the degree of exasperation aroused by the difficulty of contacting authors varies a lot from unit to unit and from department to department. And, of course, exasperation may not be one-sided: the requester who has been contacted by up to ten translators all asking the same question may well become a little impatient. The problem is that, in most cases, the individual translator has no way of being automatically informed that one of his colleagues has already asked for and obtained clarification. This is one of the problems that a bulletin board system might help to solve.
A translator will frequently be involved in activities other than translation. Some of these we shall come back to in a later section, but an attempt will be made here to indicate the potential variety of the other activities which may occupy his time. It may be worth mentioning too that translators are not always unambiguous about their feelings to other activities. Of course, other activities add interest to the working day, but several who were heavily involved in them said that they sometimes felt guilty towards their colleagues, or felt that their colleagues resented their involvement. Others feared that they might damage their own chances of promotion by seeming less productive in terms of number of pages translated.
Each unit has one or two computing correspondents, often one translator and one secretary. The correspondents serve as liaison between the unit and the computing services, and help their colleagues with computing problems where necessary. In many units, it is primarily the secretary correspondent who does most of the work. But in those units where there is a translator, correspondent or not, who is familiar with computing and comfortable with computer tools, his colleagues tend to address themselves first to him, and he may well become very involved in this kind of work. The departments also have a computing correspondent, but he is not usually a translator.
A translator may also be a terminology correspondent, helping his colleagues with terminology problems, perhaps helping to create local terminology resources and liaising with the Terminology and Language Support Services Unit.
He may also be a training correspondent for his unit, with the job of keeping himself and his colleagues informed of training possibilities.
He may also be involved with the Permanent Delegation of Translators, which in normal times meets once a month and discusses questions related to management of the Service. The delegation fulfills, for example, a watchdog function on questions of modernisation and on how they affect the staff of the Service, and plays a highly visible role in providing feedback to the decision process and expressing the translators' point of view vis-à-vis the hierarchy.
Translators are very actively involved in the evaluation of new products which are being considered for integration into the SdT environment. At the end of 1994, three commercial translator's workbench systems were being evaluated and two local terminology data base management systems.
He may also be involved in the further development of existing tools. Thus, although an experiment on using SYSTRAN for the translation of texts dealing with nuclear matters and with policy on regional development is coordinated through AGL 3, translators are in close contact with the project.
Both translators and secretaries may also be involved in Steering Committees (``Comités de Pilotage'') which are set up to follow particular projects. Each steering Committee is normally chaired by a Head of Department, and its members will include the Head of Project for the project in question. Additional members act as representatives of particular user groups, and an attempt is made to include every group which has a legitimate interest in the project. As a representative of his own constituency of users, each member is expected to consult his colleagues and to present their communal views rather than his own individual view. There is always a representative of the Permanent Delegation of Translators. There are currently four major projects, which each have a Steering Committee. The first is concerned with Follow-up, the redesign of SUIVI. The second is concerned with integration questions on the PCs, especially stream-lining the user-interface. The third deals with terminology issues, and the fourth with integrated translation tools. A fifth steering committee may soon be set up to follow work on the document server project.