Our goal is not to change or ‘improve’ meaning. It is to deliver measurable quality. This is how we manage quality control.
Translation is a transformation process. Words look completely different afterwards. They can be longer or shorter. They can be written in a totally different alphabet or in pictograms. They can be reversed from left to right to right to left or even up to down. The transformation is complete, except…it’s not. The meaning, intent, and context have to remain the same. This is the unique challenge in translating from one language to another and it can create issues if you aren’t taking certain steps to ensure that the meaning is retained. And these steps vary depending on the type of project, subject matter, and the goal of the content.
Building in preventive failsafe steps in translation workflows
Failsafe systems are processes designed to eliminate catastrophic failure if something goes wrong. For example, they can be the default to automatically close all valves in a nuclear power plant if a leak is detected. They click in after a negative event takes place. In our workflows, we take steps that might be seen as preventive failsafe steps designed to anticipate things that can cause errors in translated content. These errors can involve poor translations, translations that interpret meaning incorrectly, and translations that cause humor or offense due to cultural localization errors.
Quality control processes are not checkbox items, they are critical to success in translation
There is often lip service paid to quality control- certifications are displayed and a purchasing agent checks that off their qualification checklist. Hey, they’re ISO certified, so quality is a given, right? That’s not enough in our business. If we mess up translations, our business will suffer or even potentially fail as those errors multiply and work against our customers. So quality control processes are a subject we constantly seek to refine.
Subject matter experts, native speakers who are in-country, second translator review and edit, even machine translation
Errors are made by both humans and machines. As we build out our automation processes, there is a danger that errors can be built into a system and not be detected until it is too late. This is why the human factor will not be going away in the foreseeable future, regardless of the promise of technology breakthroughs like neural machine translation (NMT, i.e. Google Translate and others). Anyone who claims end to end automated translation without humans is simply either not telling the truth or is delivering poor translations that you could probably get by simply running them through Google.
Translator selection can be complex, especially for projects requiring multiple languages
Translator selection involves finding experienced, native speakers of the target language who preferably live in the country being targeted. This becomes important because the same language can have different standards in different countries. Mexican Spanish is different than Argentinian Spanish and both are different than continental Spanish. An Argentinian translator may not understand these differences when asked to translate content destined for Spain.
You can’t have an engineer translating consumer marketing content
The next criteria is subject matter expertise. Translating Life Sciences content can require a medical or science background with experience in regulated industries. You wouldn’t want them translating marketing content for a consumer product. Our defense and aerospace clients doing business with governments often have translator requirements dictated by government policy based on security concerns. For one client, the requirement is that the translator is a native speaker who is a US citizen and located in the continental USA. They may need security clearances and to be an engineer. When you consider multiple languages it can be very difficult to find that unique combination of expertise and skills, for each language!
Multiply all of this by two then by the number of languages: review
On top of these issues, a second translator is also used who reviews the first translator’s work and edits it for accuracy and the other criteria listed above. In the defense example, they may also need to fit these requirements. As a result of these many criteria we’ve developed preferred vendor relationships with this specialized talent, relationships that represent much of the value we bring to our clients.
Constant refinement for efficiency, quality, and turnaround time
These processes are not carved in stone. They are in a state of constant finetuning on the part of our management, project managers, and developers. There are always places where improvements can be made. Adding neural machine translation into the mix, which we do, creates another dimension to consider. This refinement process has three specific client-facing goals: cost efficiency, translation quality, and turnaround time. Ultimately, client satisfaction is the overriding priority at Language Intelligence.