Diving into SDL WorldServer, only one of many different Translation Management Systems out in the industry, I discovered the key strengths and weaknesses of a TMS; and not only that, I developed a keen eye to determining which TMS would best be recommended for individual cases. For instance, it is important to note that no TMS will ever satisfy anyone’s full range of needs. However, by categorizing the different function(s) one stresses more — such as business management, workflows, or language processing — the ranking process to deciding on a TMS will prove to be a bit smoother and less painful in the long run. Recognize the variables that play a critical role into the decision and implement that prioritization: do you perceive a heavier significance in cost, scope, quality, and/or time? For more discussion on TMS and related topics, please take a look at the following posts: Recommendations for TMS Selection; Management of Translation Crowdsourcing; and Benefits of Multidimensional Quality Metrics (MQM). In the meanwhile, read on for a quick look at a real-world simulation on how a Translation Management System can be used.
Imagine this: let’s go back a few years and assume that Apple has yet to introduce its new iPhone models, 8 and X. Apple is in need of translators and L10N professionals to localize its marketing content to resonate with the respective markets of Korea, Russia, and Spain. We have Apple jargons, various terms, and content for both translation and transcreation — and we need a translation management system to step up to the challenge. The following process and lessons learned delineate this particular scenario with Apple as our imaginary client. So humor me in this case — because believe me, one of the lessons learned will be that Apple most likely will not be a client under such circumstances. Let’s begin!
Words, words, words. Although the project proposal presented here proved to be innocuous with respect to information delivery, the lack of visuals became a major takeaway. With clients like Apple, knowledgeable in localization, or anyone else who may be comparatively less experienced in the industry, communicating important information via visual mechanisms is key. Visualization is not only crucial for easy understanding but also for expectations of the end-to-end process. For example, had we illustrated the timeline and/or the workflow of individual phases (preparation, production, and finalization), the better expectation we’d be able to provide the client step by step by having the visual representation engraved in their mind.
Below is a snippet of our quote, given that our team is localizing Apple’s iPhone 8 and iPhone X marketing home page into Spanish, Russian, and Korean. From here on, we extracted the HTML files of respective lingual pages, aligned the source and target languages through an alignment tool, and uploaded them to WorldServer to begin the translation process.
Taking the steps above, we pushed forward, little to realize that we will soon run into multiple technical issues. For example, we failed to deliver one of our promised deliverables, which were localized images. Apple has always used an image server to which all images are linked. This attribute prevents users and visitors to download the images directly from the page. In the real-world, we would need to receive the necessary assets directly from Apple’s internal system in order to localize the User Interface simulation. On the other hand, we faced complications as we progressed further into the Translation Memory (TM) creation. As seen below, Korean and Russian had significantly less matches during the alignment phase — why? To put it simply, the larger the number of alignment matches, the more translation units there are in the respective language. If it has more units, there is a higher chance that it will match with the source text, and vice versa. In this case, when we created the .tmx file, the alignment tool did a better job finding specific matches particularly with the Spanish file than any others. When we questioned why this may have taken place, the answer was quite simple: transcreation. Due to Apple’s creative take on its marketing, its source text was not translated in the literal sense. The texts below exemplify such tactic: whereas the English text states the following, take a look at the Korean text to the side. Notice the question marks — those alone indicate that a more creative spin was taken into account during the “translation” process (the last two interrogative sentences translates into “Shall we reach into our spirit animals within? How about a robot or an alien?)
Unfortunately (but fortunately — learning experience!), more issues followed. As we stepped into the official translation phase following pseudotranslation, we noticed that the pseudotranslated text had actually updated the TM for some reason and thus alternated the entire translation. Although the TM had contained the properly translated target text, it recognized the pseudotranslation as the priority text in spite of the settings we had made on WorldServer. We ended up taking the manual route afterwards, updating each individual line with the correct translation (yikes). A better solution to this problem, though, would have been to change the original workflow on WorldServer; rather than going through the generic pseudotranslation workflow, we should have added a few more steps to filter out the pseudotranslated text from the system prior to the translation phase.
Wrapping up with a few observations above, we noticed that there were some functions encoded to particular texts. This was interesting to take note of because such instances showed how the location of non-breaking spaces truly depended on the sentence structure and meaning of text that Apple wanted to convey. See above: the phrase “wireless charging” contains a non-breaking line space whereas to the right, it is broken up for visual structure. See below: the phrase “most durable glass” is broken up in the source text whereas to the right, the target Korean text maintains the phrase together as a whole in order to emphasize its message.
For an even closer look into our project, click on the links below for our deliverables per language:
Finally, we’d like to make a quick design recommendation on a crucial aspect for consideration. Multilingual typography is when two different alphabets are displayed together. As you can see from the sample image below (bottom), the baseline (indicated in the top graphic below) is not necessarily applicable to non-English text — without a careful thought into implementing a well-balanced display, the two languages might collide and hinder the overall design in target texts.
We, as localizers, must take these features into deep consideration. Precise attention to even the most seemingly minuscule detail will ultimately be what separates us as true global thinkers to local audiences.