Quality in translation is paramount in all projects, just as a production crew is essential to successfully completing a film. Translation quality works in the background as part of the backbone of production.
When translation quality slips, it quickly becomes clear how vital it is – and what could have been done to ensure a better outcome.
Maintaining a high-quality framework may be the determining factor in successfully reaching a new market, obtaining the data that you need for a market, or having a drug approved. Even though we have made advancements in the industry, the definition of quality is highly subjective.
What is translation quality?
When I started working in the industry, talking about “quality” meant dealing with customer complaints, but we have advanced so much since then. Quality is now at the center of much of what we do.
Today, both language service providers and buyers are hiring more professionals focused on localization quality.
It’s no longer a side task. It’s a full-time job.
With this shift in mentality, we are also moving from “linguistic quality” to “localization quality” because to achieve overall “quality,” we can’t simply think about whether the correct terminology is in place but rather, whether we have the proper framework in place. This includes operation processes, linguistic assets, reference material, appropriate tools, etc.
When I’m asked to define what “quality” is, the first thing I do is ask for specifics.
My follow-up question is usually, for whom? In what context? Word preference can be a quality issue in a highly creative context, but not all an issue in a medical document as long as the meaning and customer preferences are maintained.
Additionally, having a more liberal approach to adapt to idiomacy may not be accepted by the customer, but could be an acceptable translation technique when studying translation at university.
So, what is quality? In short, whatever the customer defines quality to be.
That’s why, during onboarding, it’s crucial that we define what quality is with our customers and ensure we speak the same “quality language” from the start to avoid frustrations or misunderstandings.
At TOPPAN Digital Language, we work with our customers to help them define their translation quality requirements. We bring all definitions together so that we can set their account up for success from the start instead of learning from what may be perceived as a mistake.
Unfortunately, localization is not as black and white as other industries, as perception is highly subjective. Of course, many languages have syntax, grammar, and spelling rules that need to be followed. However, customers may sometimes choose to deviate from standard language rules to fit their end goal.
Can quality be measured objectively?
Quality measurements have been evolving based on context. From The Localization Industry Standards Association (LISA) and J2450 quality metric to Multidimensional Quality Metrics (MQM), different Linguistic Quality Assurance (LQA) models have been developed and adapted to the corresponding models and needs of the industry.
There are three aspects to quality measurement:
1. Defining What Needs to Be Measured
In order to measure translation quality effectively, we first need to define the criteria that will be evaluated. Without clear baselines, conducting objective assessments becomes difficult, and interpretations may vary.
By defining the metrics upfront, we ensure consistency and create a framework that aligns with the customer’s goals. This structured approach allows us to measure performance in a way that is transparent and actionable.
2. Understanding What the Customer Deems Essential
Every customer has unique priorities when it comes to translation. Some may focus on linguistic accuracy, while others may value tone, style, or cultural nuance. By understanding these preferences early on, we can tailor our work to meet their specific expectations.
This customer-driven approach ensures that the translation aligns with their vision, ultimately improving satisfaction and fostering long-term relationships.
3. Having a Plan for Corrective Action
Despite best efforts, issues can still happen in translation. What matters most is how we handle those errors when they occur.
Having a clear plan for correcting course ensures that issues are addressed swiftly and effectively. We not only fix the mistake but also analyze the cause to prevent it from happening again, contributing to continuous improvement and greater accuracy in future translations.
Options for Measuring Quality
There have been three main quality metrics used in the industry:
1. Localization Industry Standards Association (LISA)
LISA was first introduced in the 1990s and designed mainly to assess the quality of software/hardware localization. It measured four different categories: accuracy, language, presentation, and functionality, plus three severities: minor, major, and critical. But this was last updated in 2006.
2. J-2450
Introduced in 2001, J-2450 is used mainly for the manufacturing industry. It evaluates seven primary error categories covering terminology, meaning, structure, spelling, punctuation, and completeness, plus two severities: minor and serious. This was last updated in 2005.
3. MQM-DQF
This is the latest standard to measure localization quality. This standard was created for the flexibility and adaptability needed to measure quality within translation. Measurements include style, preferences, tone, as well as the inclusion of Machine Translation. MQM is highly adaptable to the modern needs of customers and industry.
At TOPPAN Digital Language, we’ve adapted this to measure the translation quality of the services we provide, even for over-the-phone interpreting. The severities included here are not only minor, major, and critical but also preferential and kudos.
This is because we can learn as much, if not more, from what went well than from what went wrong so that we can replicate the success in future projects.
Until recently, a lot of these measurements were done through Excel files, and we needed to manually fill out the information. This was a very manual, time-consuming process that allowed us to track, analyze, trend, and action quality issues.
Then, with the introduction of data analysis tools, we were able to gain insights from the Excel files and accelerate and streamline our reporting and data analysis. Now, with the implementation of AI and LLMs and the Quality Estimations that it brings, we are at a new level of quality analysis, which is exciting to see.
AI Translation
When it comes to localization quality, the introduction of AI is an incredible time saver. It’s not about AI replacing specialists, but rather, helping us do all the manual work of filling out those Excel files and classifying the errors. Depending on the file, that could take hours, if not days, of revision, copy-pasting, and classifying.
With AI, this can now be done within minutes, and the linguist can focus on the actual analysis and look at putting action plans in place, leaving aside the manual work, and focusing on bringing the value that only a human linguist can add.
In a world where precision, context, and human insight are paramount, quality in translation and localization is a collaborative effort that defines success. By embracing advancements in AI while maintaining the irreplaceable value of human expertise, we can streamline processes and elevate outcomes.
Ultimately, the key to achieving quality lies in alignment—with customers, goals, and the ever-evolving tools that enable us to meet the diverse demands of a globalized world.