The basic purpose of a Tool...

Over the last few years computer technology has had a major impact and effect upon the age-old human activity--though relatively young business activity--of translation. Leaving aside the huge problems facing Machine Translation--as they fall outside the scope and space of this article--I intend to focus on those Translation Tools which aim at supporting, facilitating, and speeding up the work of translators. These Tools are gaining support in the field of technical translation in general, and more specifically in the field of technical translation for the Computer Industry.

If one considers that any software and hardware product requiring translation into a number of languages is periodically updated, enhanced, and upgraded, one can understand how much subsequent work on the existing translated software and documentation is needed to keep them up to date. But how do Translation Tools assist with this continuous updating process?

From Database to Memory...

Conceptually most advanced Translation Tools are alike. The basic innovation was the introduction of a translation database able to store both the original (source) and the translated (target) text. During the same translation job, if a source text is found which matches a previously translated text, the former is automatically replaced by the existing translation. That same translation database is then used to compare an old source and a new source to apply the translation for all matching text strings and leave only the new text strings untranslated.

No more retyping, and no more re-translating from scratch! Let us have a quick look at a couple of these Tools to see how they work. Of course, it is not my intention to go into a detailed description. Just a few points will be sufficient to show, that, if properly used, these Tools can really help translation.

Some examples...

Within our group we use two specific Translation Tools extensively, IBM SAA AD/Cycle Translation Manager/2 (TM/2) developed by IBM, and Text Translation Tool (TTT) developed by Dutch Seagull Business Software BV. Even though sharing the same basic concept, they differ in several significant aspects, but together they can provide a solution to a wide range of different translation needs.

IBM's Translation Manager/2...

IBM's Translation Manager/2 was initially developed--at the end of the 1980's--for internal use only. Based upon internal user feedback, it was improved and enhanced to become an external product in June 1992. TM/2 runs under Operating System/2 (OS/2) and Windows, and comprises a set of different tools working together: a text analyzer, a translation editor, a translation database called Translation Memory, and a dictionary manager. Specific file format parsers provide support for a number of different file formats, including but not limited to ASCII, SGML, Bookmaster, ISPF, RTF, Interleaf, etc., so that the Tool is able to recognize and process many different file types. The most important process in the pre-translation phase is the so-called "segmentation" performed by the text analyzer. The segment is the "measurement unit" used by TM/2 to establish the source-target relationship. We can say that a segment roughly corresponds to a normal sentence, although in some instances it can also be a single word.

Once manual translation has started, each segment is stored into the Translation Memory (database) together with its corresponding target segment (translated sentence). The source segment is then used as a "key" for looking up the corresponding translated (target) segment. If the source segment exists in the Translation Memory, i. e. if the sentence has been translated before, the corresponding target segment will be proposed as the translated version.

For subsequent updates of the same document/product, the text analyzer component is able to compare the old text and new text, and replace all exact matches (i. e., equal source segments) with the corresponding translated segments stored in the Translation Memory. At the end of the pre-translation processing (text analysis and segmentation) phase, all that could automatically be translated is already translated, and manual translation of new text can commence.

But what about those segments which, due to slight and often only cosmetic changes, cannot be considered as exact matches? Will they have to be re-translated from scratch? Not at all. The Text Analyzer can detect source segments which are similar to some extent, and store them (and the corresponding translated segments) in the Translation Memory as "fuzzy matches". During pre-translation processing these fuzzy matches are not automatically replaced with the translation. During manual translation each fuzzy match is interactively presented to the translator who will decide whether it is convenient to accept it or not. TM/2 also allows for the creation and management of one or more dictionaries where words and/or expressions can be stored. You can look up a source term stored into the dictionary, check all the translations for that term and decide which one to use in your translation. Of course, you can add, update, and remove terms from your dictionaries, and create special dictionaries for specific subjects. In short, you can manage your TM/2 dictionary system and, in general, your TM/2 translation environment in the way you consider most appropriate to meet your specific needs.

Text Translation Tool (TTT)...

While IBM's Translation Manager/2 is a flexible and comprehensive tool, a "Tool for all seasons" I dare say, Seagull's Text Translation Tool (TTT) is more limited in scope, but with a very interesting feature: it allows you to extract translatable text from a host system, download it to any PC, translate it, upload it to the host system, and merge it back into the relevant source (which optionally can also be compiled).

TTT consists of two components: TTT/PC which is the PC-based program used to perform actual translation, and the TTT extraction/update program running on the host system. TTT allows you to translate any text if it can be transferred to a PC in a TTT readable format. This Translation Tool uses parameter files to describe the format of a file containing translatable text. As long as the file containing text can be defined by a parameter file, TTT can assist in the translation.

TTT/PC runs under PC-DOS or under OS/2 with the DOS box facility. TTT, however, will only run on your PC if you plug a dongle into one of its parallel ports. The dongle has a back-end connector allowing you to connect any other device, like a printer, to that same parallel port. Just like TM/2, TTT/PC uses a translation database called Phrase Dictionary, and also a Word Dictionary. With TTT/PC, however, you can have automatic translation not only from the Phrase Dictionary but also from the Word Dictionary by pressing the appropriate key sequence.

Personal experience would suggest that this latter function be used conservatively by the translator, as single terms in one source language may have different meanings in different contexts in the target language. Therefore when using the automatic translation function for single words in message text, help text, or display screen text, the translated words should always be checked against the corresponding source words. TTT can also be used on a network without any special feature being required. The server does not even need a dongle.

Added value...

Both TM/2 and TTT offer several other functions and facilities which I cannot mention here for lack of space. However, any good Translation Tools--like the ones mentioned above--really work effectively if used by good translators. Translation Tools can add to the consistency, speed, and productivity of a translation job, but they cannot add to the "quality" of the translation. What a Translation Tool can do is "maintain" quality once it has been achieved.

A Translation Memory or Phrase Dictionary created by a quality translator is an invaluable asset, because then that same quality--alongside consistency, speed, and productivity--can be preserved for any subsequent updates of that same document/product or other similar documents/products, provided that the translator performing the updates is capable of keeping up with the original quality when translating the new text.

And some basics...

A good Translation Memory in the hands of a bad translator will never produce a quality translation. The opposite is also true. Even the best translators will not be able to make a quality translation from a bad Translation Memory, unless he/she re- translates everything from scratch! Quality is not something that can be subsequently added to a translation. Quality must progress as the translation progresses. Post-translation activities, such as proof-reading and reviewing can only fine-tune the translation work, never add quality to it, if quality is not already there. Nonetheless a skilled translator using a good translation tool will produce the best quality. Those who really want quality should never forget that the human element still plays a fundamental role and will continue to do so for a long time to come.