Translation Issues 3-98 (November 1998) by Peter Kreitmeier

Looking at escalating costs and short deadlines for foreign-language documentation, we decided over two years ago that the time had come for a hands-on study of translation tools and their practical benefits. Machine-translation systems such as Systran and Logos were not an option; instead, we directed our attention toward Translation Memory tools. We tested everything the market had to offer before finally settling on Transit from Star. There were two main reasons for our choice:

• The comprehensive package of powerful filters: As a service provider we cater for all mainstream DTP systems and word processors (Winword, FrameMaker, Interleaf, and so on). Since our primary orientation is toward Interleaf and FrameMaker we could ignore manufacturers who did not have good working filters for these packages. At that time very few companies offered an effective Interleaf filter.
  
• The price: We had to convince thirty-odd freelance translators to move over to a TMS tool, so price was a frontline issue. Star won out over many rival companies by offering "light" licenses at an attractive price. These licenses, as the name suggests, are stripped of certain functions such as tc-forum/export filtering and segmentation. This is not necessarily a disadvantage for freelance translators if the service provider prepares the translation projects, as we do.
  

• Transit's user interface could scarcely be termed intuitive: some features are throwbacks to the dark days of DOS. File management suffers accordingly (Transit has yet to achieve the 32-bit level on the evolutionary tree).
  
• The software is sophisticated but the user documentation is woefully inadequate. There is no Getting Started section and users progress rapidly downward through disappointment to discouragement to nail-biting frustration as the User Guide consistently fails to explain major concepts in the Transit environment.
  
• Project preparation is cumbersome. The value of thorough preparation is undeniable, however, because this phase (text segmentation, choice of reference documents) has a direct and powerful impact on output quality. Hidden text in online help documents, for example, often has to be tagged to ensure that it is not translated, but isolating these strings can be difficult. Care also has to be taken in defining rulesets for handling abbreviations. Many of these problems would not arise if writers take the time to consider how quirky formatting can hamper translation.
  
• Translators see only a few language-pair segments on the screen at a time. The information content of graphics and cross-references, for example, is left behind in the original document. This drawback is inherent to nearly all Translation Manager systems. Forced by the tool to adopt a sentence-by-sentence orientation, translators who are not content to work at one remove from the subject matter find themselves switching back and forth between the TM user interface and a paper or DTP original.
  
• The data has to be managed in two systems (Transit and the DTP program). Once the files have been exported from Transit for proofing, subsequent changes can entail twice the work to ensure that the Transit reference library accurately reflects the final DTP product. The TMS tool is unable to cope with corrections of certain kinds. This is true especially of changes to tables and graphics. Errors tend to crop up again in the next version and refuse to go away as the document progresses through each successive release.
  

• Updates are much easier to handle. There is no need to make line-by-line comparisons to pinpoint changes before a new document is generated. These checks are highly automated and error-free.
  
• All the files in a project can be loaded into Global View, an excellent platform for quality-oriented functions (such as search and replace).
  
• Online terminology does away with wearisome searches through serried ranks of well-thumbed dictionaries. The automatic terminology check is ideal for monitoring consistency – assuming that someone of a lexicographical bent has compiled a dictionary.
  
• Post-editing is much less time-consuming, especially with Interleaf and FrameMaker documents. This, too, is subject to a proviso – the source documents have to be adequately formatted.
  
• Translation quality improves, because departures from one-to-one translation accuracy show up more readily in workflows.
  

"transline Deutschland - Übersetzungsdienst für technische Übersetzung"