MAXIMIZING HUMAN TRANSLATION: The Power and Limitations of Computer-Assisted Translation Tools
So what is Computer-Assisted Translation? Is this the same as Google Translate? Are human translators even looking at my documentation? Don’t you just press the “French button”? Yes, we have heard all of these questions. Here’s a glance of the translation environment:
Computer-Assisted Translation software combines very valuable functions to serve HUMAN translators:
- Segmentation: When a translator (or Project Manager) imports a source document into a CAT tool, the software breaks the text into segments, which might be sentences, paragraphs, headings, or bullet lists. These segments are usually delimited by punctuation marks (period, comma, carriage return). The theory is that bite-sized segments might be repeated from one chapter to the next or one document to the next.
- Translation memory (TM) is the database that stores segments as these are translated by human translators. The translation memory stores the original sentence and its matching translation in bilingual pairs called translation units.
- Glossary or TermBase (TB) is another bilingual database like the memory, just with bilingual pairs of individual terms, A term might be a single word like camera or it might be a set phrase like thermal imaging.
- Analysis: Project Managers often use this feature of CAT tools to estimate how many segments in a document are new text versus how many segments already exists in the Translation Memory.
- Filtering: The software parses the source document and locks out formatting or programming codes that should not be altered by (nor distract) the translator as he works on translating the actual words. Filtering prevents XML tags (for example) from accidentally being corrupted during translation. It also converts specialized software types (e.g., Adobe Framemaker, PHP) into a common format that translators can efficiently handle without advanced desktop publishing or programming skills. At the end of translation, the CAT tool flows the translation back into the original format type. At this point, an advanced desktop publisher or graphic designer can apply the final polish to the translated layout, such as adjusting pagination and kerning to accommodate a translation that occupies more space than the original document.
- Security Note: Although some CAT tools do also connect out to automated machine translation software like Google Translate, Verbio strongly advises against anyone using this feature because Google Translate captures and stores all phrases entered into it. This poses a serious liability for security-minded businesses. To be frank, Google Translate results are often more amusing than serious translations.
Effectively CAT tools are just that: tools that assist a human translator by:
- Converting a source file into a usable format for translation by humans.
- Notifying a human translator when a segment exists in the Translation Memory and showing the translator how the segment was translated previously. Human translator always has the option to modify or completely overwrite the previous translation based on the new context.
- Offering human translator a database of specialized terminology directly within the software environment and as a supplement to the physical bookshelf of technical dictionaries.
Limitations of Computer-Assisted Translation software
Although CAT tools can significantly reduce the translation time of large translation projects, they also present some limitations that are worth keeping in mind.
- When text is extracted from the source document into a plain bilingual table, translators often lose the context of the surrounding paragraphs and graphics. Therefore, it is helpful for translators to have a PDF of the source document as a reference in choosing the right words for the right passages.
- The format of the final translation will also look somewhat different from the original after the translation is placed back into the original layout. Again, context is more readily apparent in the final layout. For these two reasons, translations need to undergo a measure of document layout work and a thorough Quality Assurance pass on the final layout, in order to ensure that everything reads as expected.
- CAT tools are highly useful for repetitive types of texts, such as technical manuals, employee handbooks, or software interfaces. When text is not repetitive (e.g., marketing or literary), one loses some of the advantageous features of being able to leverage segments stored in the memory. Notwithstanding, repetition is a pattern that can emerge within a single document or across a series of documents. So the relevant use of translation memory is something to be considered carefully with your translation team.