Translation quality assurance (TQA) - best practices for software localization

Translation quality assurance basics

Translation and localization might differ, but whichever process is in question - they should both end in quality assurance. While QA is often an afterthought, it can be a massive saviour of resources (time & costs) and ensure products are delivered to the target audience and received as intended.

You can count on poorly translated content to raise distrust among your customers, while poor localization can downright bring misunderstandings, confusion, or even offense. TQA is the last step towards assuring a well-adapted product that speaks to your customers through content and cultural lenses.

What is translation quality?

Translation quality is not easy to determine, especially if we consider how many different stakeholders are involved in the investment in the product, participate in the translation, or those who use it.

Quality translation might mean your translation of optimal accuracy, delivered within (or earlier than) the deadline. For a translator, quality translation means the right localized content. But in the end, the user makes the call, and her behavior will tell us if the quality matches the expectations. In other words: if people don't use your product, you might want to re-evaluate your copy & translation quality.

The variations are endless, but what all quality translations have as a standard solution is an intersection of all these elements and perspectives as much as possible harmony. High-quality translation and localization will provide the end user with an authentic product experience by maintaining linguistic and cultural standards within the project scope.

Translation quality process

A good translation quality process has a couple of different stages that span across and go in line with the development process. We are inclined to believe that the process looks something like this:

1. Internationalization
2. Localization
3. Quality assurance

While many companies opt for this approach, a couple of hidden bums are on that road. Here's what we've discovered can save you a lot of time and headache along the way:

• Pre-localization checks: source content verification

Often, issues during localization come from intangible or incorrect source text. If you are fortunate and you're doing your localization with the help of a very skilled linguistic professional, they can catch the mistake and correct it. It's more likely, however, that the translator will get confused and have to check for the meaning of the source text, which results in delays and corrections on both sides. The worst-case scenario is if you rely solely on machine translation and have yet to assure the source content accuracy beforehand. This scenario might land you and your product with a place on a list called "The best translations fail" or similar.

• Functional testing: grammar, placeholders, and commas

Functional testing comes right after the localization, or "first draft translations." Functional testing targets inconsistencies such as spell checking, grammar, presence of placeholders, formatting, multiple spaces, repeated words, plurals, and consistency of upper and lower case letters.

Functional testing is often automated or implemented through machine engines and CAT tools. This speeds up the process significantly but may lead to missed or even false checks. The missed quality checks are mistakes the computer or tool cannot see. False checks are checks the computer detects, although they make sense. For example, think of Microsoft Word and a spell checker that lights up whenever you type an unusual name.

• Linguistic testing: meaning behind the content

Linguistic testing is usually the most demanding step of the process, where you will probably want to have language professionals on your team. Linguistic testing ensures the content's meaning and intention are properly conveyed in the localized, target content.

Additionally, in this stage, you want to ensure complete consistency in your localized product. For example, if several translators worked on your project, they may have used different terms to translate the same thing. In this case, translation is usually correct from the functional point of view - not accurate for your specific project. Aside from professional translators, you need a term base or translation memory for this step.

Term bases and translation memories are collections of terms and preferred translations that suit your project. This can be the perfect guideline for your translators. If you have assured it at the beginning of the project, the translators or machine translation engines will probably be able to implement it from the very start of the process. This means less editing, fewer corrections in the TQA process, and money and time saved.

• Visual testing: the feel-good check

Visual testing is the perfect stage to bring in native speakers and beta testers. Once content consistency and accuracy have been verified in the previous steps, you can get valuable feedback from would-be users. They can give you input on the layout, the amount of content, and how user-friendly it is.

This is especially relevant if you are localizing for a market that is culturally very different from the one for which the source product was developed. Visual testing is crucial if you translate into Right-to-Left languages or vice versa. The essential value of the visual testing of your translation is that it evaluates the "feel-good factor" or the user experience - how well adapted your localized product is to your target users and how close to the original creation of the product is.

Ideal translation quality assurance team

When you think of an Ideal Translation Quality Assurance Team, your mind probably goes to an image of some multilingual version of Shakespeare. Sadly, that wouldn't be the best option for two reasons: firstly, because Shakespeare was notorious for making up new words when he couldn't find a fitting one, and secondly, well, he's currently indisposed. There are, however, skills and people you need on your side in this adventure:

• Engineers & Developers - people with product knowledge are crucial. They are the ones that are the backbone for translators and language professionals. They know what each section is supposed to do and what should be the result of each action.
• Translators - make the words say what you want them to speak in as many languages as you wish. Finding reliable, trustworthy translators with a good track record is priceless. A good translator will cut your QA process significantly, and an excellent translator will give you the illusion that you don't even need a QA process.
• Editors - a perfect editor is a language professional, a native speaker, and understands your product from the user's point of view.
• QA professionals - QA testers will understand the process and deliver you the results you hope - they have the knowledge and experience. If you don't have access to QAs or your QA is busy on other projects, ensure your translation management platform streamlines the process for other team members as much as possible. This will not compensate for the issue entirely but will save you time and avoid trouble.
• Beta users - Native speakers who can do the visual testing and don't necessarily have to be software engineers or language professionals. They can give you the paramount insight if your localized product can enter the race for market domination or if you're on the slippery slope of mistranslation.

TQA Approaches

There are different approaches to translation and localization quality assurance. The division can be made by focusing on which stage of the process QA happens, which tools are used, or simply by the ones performing the QA. In general, we can easily see the two main approaches:

1. Quality Assurance systems & tools. These quality checks - translation quality management systems are automated and usually go hand-in-hand with Machine Translation. This solution is perfect for acceleration, especially in functional testing - spell checks, placeholders, formatting - all of this can be reviewed with a high success rate using CAT tools. With a high level of success, good QA can help you verify the consistency of term usage and employ a verified Translation Memory.

The usual downside is that most TQA tools operate on a bilingual level. This means that it's possible to find inconsistencies between the source and target language, but if there are more significant issues that need to be resolved, most TQA tools won't help you tackle this issue.

Since Lingohub founders encountered this issue so often in their localization experience, one of the main features they wanted for the Lingohub app was the possibility of editing multiple translations. This sped up the localization process by streamlining the editing and functional testing using Lingohub's Side Panel.

It's fair to say that our biased love towards this Lingohub feature came from years of navigating between 5 or more translators and translations when only one placeholder had to be changed. Our life is easier for it, and we hope it will make your life easier.

2. Native language professionals. Native speakers who are native language professionals are the dream of any developer diving into localization. Using language professionals brings more accuracy to your localization than basic Machine Translation. They will be able to catch logical mistakes and inconsistencies and bring higher value to your overall UX.

However, even the best language professionals are human, and CAT tools can only solve some things. Relying entirely on human QA testers means your process will be slower than automated QA tools. If you have fixed deadlines and need a quick turnaround, you might be in a situation where you have to sacrifice either time or quality.

3. The hybrid quality assurance approach. Choosing to sacrifice either product delivery time or the quality of the localized product is usually a nightmare for any Product owner and manager. We don't believe in this compromise, so the Hybrid Quality Assurance Approach works best for us. What does it mean?

Based on the principles of Continuous Translation, a hybrid approach synchronizes as much as possible different stages of QA testing (source content verification, functional and linguistic testing) by applying simultaneously both QA tools and language professionals.

• Possibility of editing source text quickly at any time.
• Term Bases and Translation Memory are actively used by the translators in the localization process, not just for last-phase review.
• Quality Checks are displayed in real-time, so translators can immediately adapt the content in the localization process.
• Placeholders are detected in a real time.
• Access to functional testing across different target languages in one window.
• History of changes at a glance and possibly reverting to previous solutions.
• Use translators as beta testers - we recommend providing translators with access to either beta versions of your software or at least visuals and action context. This way, the translators get to provide translation that aligns with the desired user experience.

The hybrid approach to TQA is quick and flexible, allowing acceleration of the translation & localization process without sacrificing the quality. The downside is that a messy translation management platform might confuse team members. Simple features and intelligent solutions - that's the type of tools you want for your team.

TQA future

No matter how good, at the moment, QA software is still just a machine that can overlook the nuances of human language. Just because mistakes are not detected - doesn't mean they aren't there. On the other hand, even the best human language professionals are still just human and cannot compete with machines regarding translation speed.

With that in mind, it's fair to say that there is great potential for using Artificial Intelligence in the translation quality assurance processes. The technology is going full speed ahead, and we at Lingohub are doing our best to be part of the race with our Linguist project. In the meantime, we put our trust and faith into the hybrid quality assurance approach.