Takeaway: Reach for the STTARS

undertake goes straight into the heart of quality: the source language. We should be looking at the source content and trying to anticipate questions and problems for translation. These may cause confusion, more work or rework if they are asked later in the process, such as during translation, for example. Source analysis can be intensively computer-assisted. Many sources of confusion can be systematically extracted from the source, then clarified and smoothly translated into 30 languages later. It is possible, for example, to use tools and create rules to find all the acronyms in the source content. Acronyms are a pain to translate. So let’s clarify them! And let’s do it before translation. The questions that we ask during source analysis are questions that will not be asked later 30 different times for 30 languages.

T is for terminology. We should extract terminology terms before translation, and this can also be a computer-assisted task. Terminology is a big money maker, and we might be missing out. It actually can make things better, cheaper and faster. Think of a term that appears 1,800 times in a software product and can be inserted with a keystroke and translated consistently 1,800 times. Feel the savings and quality? We can also collect terms that are really important for the success of a product in a certain country — or actually in all countries. They can be researched and explained, then translated and validated by an expert. Then all 30 languages will use a high-quality translation for these terms consistently throughout the entire project — and future projects. And after the translation is done, we should once more use tools to check if the terminology was actually followed.

T is for translation. It ruins the acronym a bit, but we have to translate. This could be computer-aided translation or machine translation, and there are various technology tools to help us.

A is for automation. After we translate, we should start revising the translation, relying heavily on automation instead of human work. There are many automated quality checks that can be performed using tools and rules to find specific issues. Let’s not forget the oldest automated check, the spellchecker! Also, if the source segment equals the target segment, there is a chance that the segment was not translated. If the same source segment is translated in two different ways, there is a chance of an inconsistency. We all know that there is a human validating the errors, so the term automated does not mean “non-human,” but it does emphasize the automated nature of finding the errors as opposed to reading the whole text.

R is for revision. The EN 15038 standard defines revision as the task where someone looks at the source and target to find errors. For many people, this is also what the E in TEP stands for. So, this task is the familiar human review of the translation. However, it will be more efficient if extensive automated checks eliminate a variety of errors, and only the errors that can be caught by humans, such as a misunderstanding of the source language, are left.

S is for signoff. A review in the final format that will be seen by the end user of the translation is not to be underestimated. I have a printer manual that contains Danish segments among the Portuguese translation. This is the kind of error that we want to prevent with this task. However, this should not be a linguistic task in the sense that you should not be fully reading the content. The time to find these kinds of errors has passed. Here you’re trying to find major errors without reading every word.