Trados: smarter translation software

Translator’s Workbench II is a new integrated translation package from Trados. The question is, will translators be ready and willing to make the move?

Trados, the Stuttgart developers of the MultiTerm terminology management program for PC/compatibles, is now shipping the Translator’s Workbench II package it demonstrated at cebit earlier this year. The company has also announced a marketing agreement with Eurolux, familiar in the translation world as the European distributor of MTX/Mercury (See  sidebar ). These moves, together with the Microsoft Windows version of MultiTerm due shortly, demonstrate that Trados is working hard to establish itself as the preeminent supplier of translation software. Trados has competition, though, not least of all IBM, which has also recently released a translation sofare package (see below).

Yet another text editor
Translator’s Workbench II comprises the TW II Editor (formally TED) and MultiTerm 2. With MultiTerm, Trados addresses the matter of consistency, in particular the use of consistent terminology among a group of translators or across multiple projects. For those unfamiliar with MultiTerm, it is a terminology management program which takes the innovative approach of storing all data in a single, freely-structured database with entries classified by user- defined attributes.
    With the TW Editor, Trados adds facilities which increase translator throughput by automating the translation of repetitive material. Trados takes the somewhat controversial approach of requiring translators to forsake their familiar wordprocessing software for a new editor. While the Trados Editor is an acceptable text editor, it does not offer the exhaustive (some say exhausting) facilities of a full-blown package like WordPerfect. Whether this remains a contentious issue or fades into oblivion with the arrival of Windows-based wordprocessors having consistent interfaces remains to be seen.
    By means of its pull-down menus, you can configure the Editor to automate your translation tasks in a variety of ways. The Active Terminology Recognition feature, for example, will automatically scan the current sentence for terms stored in the MultiTerm database and highlight them. Upon pressing F6, a small window pops up to display the translation(s) of the term in the target language. You can select a target term and paste it into your document with Alt-A. With the Steno function, you can assign long strings of text to an abbreviation; it will then expand the abbreviation to the full string automatically when you type it.
    For users working in text files containing formatting tags for such packages as Ventura Publisher, a Protect Tag feature is available which automatically recognizes the tag names in the user-definable configuration file, “locking” them and changing the color. Provided the required preparations are made, this could potentially save translators who have to work within the confines of specific typesetting formats a lot of grief. It also has intriguing potential for use with SGML in the future..
    The centerpiece of TW II is surely the Translation Memory feature. After switching this mode on and indicating the .tmb file (a translation database), you press Ctrl-T to activate it. The program will scan through your source text, segment by segment (usually a sentence), and search for a similar entry in the .tmb file. It uses sophisticated fuzzy match algorithms to try to match source and target texts. These algorithms morphologically reduce significant words, index them, and try to parse a sentence syntactically. (As you would expect, it is a TW Editor currently supports German, English, French, Italian, and Spanish.) If a one-hundred percent match is found, the target text will be automatically pasted into the editor, replacing the source text. If similar but not identical text is found, a window will be displayed showing both original and new sentences, with deviations between the two texts (text added, deleted, or moved) indicated by various colors, similar to the DocuComp system. If there is no match for a given sentence in the tm database, it recognizes it as a new sentence and prompts you to enter a translation. A powerful feature of the memory is its ability to substitute terms automatically which it recognizes from the tm database..
    A taste of pre-cooked translation memory can be had by loading some of the test files supplied by Trados (with plenty of 100% hits), activating it, and watching it zip though the text, replacing the English text with German. You might be tempted to wonder: who needs Systran? Achieving the same effect in a real- world situation, however, will take considerable time and effort. Translating with both source and target text on screen may also require some adjustment – and a good monitor.
    To be able to put the translation memory feature to work, it should be obvious by now that you need to have the source text in binary form. Fortunately, many if not most translation companies have been able to convince their customers to supply the texts to be translated on floppy-disks. Less obvious but not less vital is that it will require significant initiative and energy to build a useful translation memory. While you can begin using translation memory immediately, it could be a matter of weeks or months before clear productivity gains are achieved, and then only if you are translating suitable material. Trados’ Jochen Hummel acknowledges this, saying, “the price of the program reflects the fact that it is a sophisticated, specialized tool.” He suggests that it will take technical skill and a strict implementation plan to adopt TW II to specific situations and working methods. This implies either internal technical support or external consultants.
    While it is not likely that many independent translators will be able to justify a system like TW II, Hummel does mention one British translator – a long-time TED and MultiTerm customer – who was quick to embrace TW II and has already had a program written to generate a translation memory database from several years’ worth of translations. The key here was the availability of both the source and target texts for a large body of material translated for a specific customer along narrowly defined guidelines.

Who is Trados?
The Trados software was developed by Jochen Hummel and Iko Knyphausen, the founders of Trados. They have their origins in the translation world and, in fact, their translation subsidiary, INK/Deutschland, is, as the name implies, a member of the INK/International network. According to Hummel, Trados’s development efforts are now self-sustaining, although he says this is due partly to their keeping their operation lean. Trados consists of the two developers, their secretary, and a computational linguist at the University of Stuttgart.
    Discussing Trados’s customer base, Hummel explains that the typical MultiTerm site is five or ten ne networked users who form the technical writing and translation department of a medium-sized, export-oriented company. Of these users, two or three might be terminologists who are responsible for building and maintaining the terminology databases for the group alongside their other activities. Hummel believes that Trados has barely scratched the surface of this market; in Germany alone there are probably thousands of such businesses which might benefit from terminology management software.
    According to Hummel, Trados will shortly be introducing a Microsoft Windows version of MultiTerm which will support graphic images and will adhere more closely to the iso terminology format (see page 9). He points out that Windows could be the key to more widespread use of terminology management software and explains why. “Ideally, you want everyone in your organization to have instant access to your terminology database. Not all users, however, can use TSR software because of its memory consumption or other reasons. But what if the program was simply an icon on everyone’s Windows’ Desktop? If you wanted to look something up, you could just click it. Even if you only use it once or twice a month, it wouldn’t get in your way.”.
    Windows brings with it other advantages too, says Hummel. While the interaction between a memory-resident TSR program and wordprocessing package being used takes place, technically speaking, on a rather low level – keystrokes are stuffed into the keyboard buffer by writing directly to memory – programs running concurrently under Windows can interact in a far more sophisticated way. According to Hummel, the imminent Windows version of MultiTerm will use the Dynamic Data Exchange (DDE) facilities of Windows and macros written in Word For Windows’ WordBasic to offer the equivalent of the look-up and paste functions found in the DOS version. “It’s a far more powerful interface,” comments Hummel. “With dde, you can pass complex data structures back and forth between programs. We could, for example, generate a table in Word containing both the source and target terms.”.
    While Trados has not announced as much, it seems highly likely that the company will develop a Windows version of its Workbench. Hummel acknowledges that TW II begs a fully windowed, graphical user interface. At that point, the two developers might find it logical to offer separate windows for the source and target texts, rather like the old Alpnet TSS software, which in many ways is the spiritual forbearer of much of today’s translation software. A further speculation: the background processes inherent in TW II would seem to find the perfect match in a multi-threaded operating system like OS/2; what kind of services might then be possible? It might also be conceivable – if not extremely desirable – to implement the translation services as independent programs which use interprocess communication, thereby allowing users to access them from whatever wordprocessor they use..
    Over the course of coming months and years, it should become clear to what degree Trados, IBM, and others have succeeded in identifying and addressing translators’ needs. Do packages like TW II go far enough in adapting to the way translators work? Are translators otherwise willing to adapt to the software? Do such packages mark a transition from the use of general “business” software by translators to software written specifically for them? Be that as it may, the challenge remains formidable: to enhance, not inhibit, the vast amount of knowledge that professional translators have stored in the ultimate online device – their brain.

Prices: MultiTerm DM1800, Translators Workbench II (incl. MultiTerm) DM4800

Trados GmBH, Rotebühlstraß;e 87, 7000 Stuttgart 1, Germany Tel +49 711 627 068, Fax +49 711 616 745 Email 100042,66@compuserve.com

(See  sidebar that accompanied this article)