By Mike Dillinger, PhD, Translation Optimization Partners (with contributions from Laurie Gerber)
Previously published in ClientSide NewsOptimizing the translation process has two basic components: improving the writers’ source texts and improving the translators’ process. For the moment, we’ll focus on the writer’s job.
Dear Translator: Please remember that most writers never had any training at all about translation and usually know one lonely language. Many of them can only rely on the limited writing advice that they got in school. They’re never aware of how they can make life hellish for translators and for international readers. So, don’t blame them; help them out. Pass this list on to them and discuss it until they understand.
Dear Writer: Become a hero among your company’s translators. Every improvement you make is multiplied by the number of grateful translators you help. This list is for you.
Pothole #1: Thinking that your original source text is the final product
Translation takes a long time, so it is very important to budget enough time for converting files, translation itself, desktop publishing, re-doing screenshots, interface localization, etc. “Express” translations are done with shorter-than-usual deadlines and cost two or three times more than regularly scheduled translation jobs. We’ve come across translation vendors who very happily report that the majority of their work is from writers who couldn’t get their schedules organized. Writers who are pressed for time also hit many more of the potholes listed below and that makes translation even bumpier than usual.
Similar problems show up when we use machine translation (MT) to speed up translation. Shorter deadlines mean that there’s less time available to tweak the MT engine for a given job. So, in this case MT makes more mistakes and requires more time for editing.
Remember that your final product is the full set of documents, in the source language and in all of the target languages.
Pothole #2: Assuming that your files will work everywhereTranslators are generally well-organized folks and focus on optimizing their efforts. So, the first thing they do is take your original files and run them through translation memory (TM) software to see if any of your sentences were translated before. That’s easier said than done. If your files are in some Adobe FrameMaker, Microsoft Word, or other proprietary format (i.e., not an open standard like HTML or XML) then translators will have to convert them. According to Murphy’s Law, your specific version of the authoring software will not be completely compatible with the specific version of the translators’ TM software. So, translators have to check and fix your converted files by hand, one by one. Remember those extra line breaks that you added to improve the formatting? Well, they really interfere with converting your files. The translators will charge you for all that unpleasant work by the hour, often for each language – on top of the price for translation itself.
Using the wrong file formats will make translation slower, more expensive, and more error-prone. And this is work that the translators will have to do all over, again and again, for the next versions of your documents.
Pothole #3: Using screenshots for eye candy
Many translators specialize in translating manuals for software products, which often contain lots of screenshots. And the screenshots look really nice. But screenshots are very difficult to translate! For one thing, it’s hard for translators to figure out exactly how to see the same exact screen so they can take a new screenshot in the target language. This is doubly true for error messages, which are hard to produce on-demand in any language. Often, localization of the software itself hasn’t even been finished yet (usually, a different team is working on that) so the translators working on the manuals don’t have the target-language product in front of them. Sometimes they have to edit the image by hand to cut-and-paste and draw in the translated words – even if the words have nothing to do with what the screenshot is supposed to illustrate. So, the translators have to painstakingly re-create your screenshots by hand, one by one. They’ll charge you for that by the hour, for each language – on top of the price for translation itself.
Think again: Do you really need all those screenshots? Using too many screenshots will make translation slower, more expensive, and more error-prone.
[Editor’s note: No, you really don’t need all those screen shots. In fact, you may better serve your users by providing video snippets of how the feature actually works; that too, will involve translation and localization effort. But, if you’re going to pay to translate and localize, why not pay for something the end-users actually find valuable?]
Pothole #4: Thinking that your page layout will look the same in every language
Translations in many European languages take up about 20% more space on the page and some Asian languages take up less space than English. If writers don’t leave a lot of white space in the original, then the translators have problems. If translated content spills over to another page, then either the layout has to be re-done by hand for each language, or the pages have to be re-numbered (everywhere!), or both. Once again, translators will charge you for that by the hour, for each language – on top of the price for translation itself. And this is work that they’ll have to do all over, again and again, for the next versions of your documents.
Pothole #5: Trying to make your writing “interesting”
In high school, our English teachers wanted us to produce more “interesting” writing. They wanted us to use different kinds of sentences, phrase things in different ways, and even play with words. And this is how we’ve seen writing taught in other languages, too. This is a fine approach for producing literature that native speakers will read.However, this approach makes life very difficult for translators, in two different ways. On the one hand, translators usually read English as a second language, so they don’t know as many of the nuances as a professional writer does. That means that the translators sometimes think that different phrasing has different meaning and they often have to sweat to render the (probably unimportant) difference in the target language. On the other hand, phrasing the same information in different ways means that you also have to pay for it several times. Remember the TM software we mentioned above? If you stick to the same phrasing for the same information, the software will see that and you’ll get the second and later translations of that information (almost) for free.
Similar things happen with machine translation. MT can translate many kinds of sentences very well. However, Murphy’s Law strikes when writers use varied phrasing: It’s much more likely that they’ll produce sentence types that machine translation simply can’t handle. That in turn means more editing effort and longer delays in translation.
Pothole #6: Not being picky enough with words
Picky is good. Picky writers choose their words very carefully, with the reader in mind. Translators are readers, too, and they have to understand what you wrote to translate it correctly. But many writers spend too much time talking to engineers and to each other, so in-house jargon, invented words, and unusual (for the rest of us) technical terms often show up in client-facing documents. You may be used to “noun stacks” like strategic application development productivity or KX Manager Device System Information Configuration Backup, but your translators and end users won’t be able to digest them easily (see more discussion here).In pothole #5, we saw how even seemingly innocent synonyms are problematic. Even so, in most documentation projects, 75% of the words are used only once or twice. For translators, this means several things. Translation memory software doesn’t find as many matches, because writers vary their phrasing even though the information is the same. Translators have to translate the same information several times, simply because it was phrased with different words. Machine translation and human translators are less likely to know all of the words, because they have smaller vocabularies than a professional writer does. That means more terminology research and more mistaken translations.
Pothole #7: Going soft on verbs
Things get tense for translators with English verbs. Professional writers are often comfortable with compound verb forms like may have been being installed, but this is enough to tie anyone else’s head in knots. Words ending in –ing are also a big problem. These –ing words can be nouns or verbs, and readers can understand them in multiple ways. One favorite example is a book title by Rachael Ray: Cooking Rocks. It might be about: how to cook rocks (like cooking potatoes), rocks that you can cook with (like cooking utensils), the kind of rocks that you can cook (like cooking potatoes, not planting potatoes), or it may be something else entirely (like cooking is cool). If your translators or readers aren’t very familiar with your product, they can easily get misled by this kind of writing. Machine translation systems are forced to guess the right translation. Going soft on verbs can lead to lots of errors.
Pothole #8: Speaking to the wrong audienceWriters are often hired to explain how to use a product that readers aren’t familiar with. So, managing assumptions about what readers know about the product is an important issue. But writers are so overly familiar with the product that they often assume that readers are familiar with it, too. This shows up as abbreviations and acronyms that newbies don’t know, as technical, rare, or unfamiliar words, or as sentences that can have multiple meanings. These are all perfectly clear to the writer, but often mysterious to the reader. Even tiny things like definite articles (the) create problems: they show up when the writer assumes that the reader knows something. One example is: Once you save a record, the fields become read-only if the record’s approved. This is fine if the reader knows that records have fields, but there was no mention of fields anywhere in the same chapter as this sentence.
People only understand and translate easily when there’s enough information in a sentence that they don’t have to use detailed topic knowledge to fill in the details.
Pothole #9: Talking over their heads
Pothole #8 was about what readers know about the product. Pothole #9 is about what readers know about English. Most people don’t read as well as professional writers do. In fact, the average reading level for native English speakers in the US is about 7th grade. On top of that, translators generally work from their weaker language (English) into their native language. Long, complex sentences not only take much longer for mere mortals to understand, but lead to more comprehension and translation errors. One added benefit of shorter sentences is that they are easier for writers to reuse in different contexts and are easier to match in translation memory. Translating short sentences is faster, cheaper, and more consistent. Clients prefer them in English, too.
Pothole #10: Being too hip for the audience
As mentioned in Pothole #5, paraphrases, word play, slang, and metaphors work great for literature. Marketing copywriters often think of their job as coming up with quirky, catchy phrasing everywhere. All too often, though, the information that’s most relevant for the reader is hidden or missing. A simple, direct statement of how the client benefits from a product is a refreshing change that’s easy to translate. Most manuals focus on the geeky details of a product that the majority of readers couldn’t care less about. Simple, direct information about how to actually get something useful done (open file doesn’t count) is surprisingly rare. This is the issue of relevance.Relevance is a huge issue for writers, and one that really hasn’t been looked at very carefully. Some companies, like Intel and Microsoft, track readers’ ratings of how useful or helpful the company’s content is. The results are horrifying: Only 25 – 40% of the support content that readers can find is actually helpful for them. On top of that, there’s probably another 20% of the content that readers never find, and this is frequently because the author used one term but the reader used a different, less technical synonym. If similar numbers hold true for other kinds of product information, then we’re writing, revising, translating, and publishing millions and millions of words that no one cares about. Simple, relevant content makes translation much faster, much cheaper, and much more accurate.
Translation speed, quality, and cost depend to an enormous extent on the quality of the original documents. Every single choice of wording, sentence structure, style, and content influences how easy or hard a document is to translate. Writers, help your readers and become a hero among your company’s translators. Every improvement that you make is multiplied by the number of grateful readers whose lives you touch.
About the Author
Mike Dillinger has built and deployed language technologies for over a dozen years. He has an especially broad perspective on industrial best practices and their role in building an integrated multilingual content supply chain. This comes from his extensive international experience as a management consultant, researcher, technical writer, editor, translator, and developer of translation software. Mike has special expertise in strategic planning of global content.
He has worked with content and translation groups at Apple, eBay, HP, Motorola, Cisco, MRO Software (now a part of IBM), Borland, Raritan, the Marshall Center for Security Studies, and others.
You can reach Mike via email at mike@translationOptimization.com.