Article
March 3, 2025
Over the past ten years, content creation and distribution have reached unprecedented scales. This expansion, driven by the increasing global nature of modern business, has brought both new opportunities and significant challenges in maintaining quality across languages and cultures.
The rise of AI technologies, particularly in translation and localization, has created a complex landscape of possibilities and pitfalls. While AI has been part of the translation industry longer than most - with the original Transformer model designed specifically for machine translation - today’s environment of heightened AI awareness has created new dynamics.
Many vendors have leaned into the current AI hype cycle, emphasising large language models and automated solutions that promise quick, cheap translations. However, for organisations with high-value content, this approach often falls short of delivering the required quality and precision.
For organisations with high-value content, we need to think differently. We define ‘precision content’ as materials that require meticulous attention to linguistic nuance, cultural context, and technical accuracy during translation and localization. This includes content where brand voice is critical, specialised domain knowledge is essential, or where the content has significant longevity or reach.
As businesses worldwide reassess their content strategies in light of these technological advances, we see an opportunity to reframe the discussion. Rather than focusing solely on cost reduction through automation, we believe this is the moment to examine how we can leverage technology, including AI, to deliver precision content when working globally.
For many outside the translation and localization industry, translation is often perceived as a straightforward process of swapping words between languages. This perception has only been reinforced by the emergence of Neural Machine Translation and Large Language Models (LLMs), which make it remarkably easy to transform text from one language to another with a simple API call.
Whilst this approach can be suitable for certain types of content – such as chat messages or ephemeral communications – it falls short when precision and quality are paramount. High-value content that reaches wide audiences or serves as evergreen material requires a more nuanced approach.
So what does this nuanced approach actually entail? Let’s examine the complex reality of what truly goes into creating precision content across languages and cultures.
Professional translation and high-quality content localization demand consideration of multiple factors that go far beyond simple word replacement. These include:
Effective localization requires careful adaptation of examples, brands, and visual elements to resonate with target audiences. For instance, the colour of medical professionals’ uniforms varies significantly across countries, and marketing examples that work in one culture might fall flat or even offend in another.
Many real-world content types require the handling and preservation of complex formatting – from print-ready formats like InDesign and FrameMaker to structured content in XML-based formats such as DITA. Maintaining this structural integrity whilst adapting the content is crucial for ensuring the applications or documents remain functional as well as the effort of professional presentation is maintained. No one wants to re-do work just to use a translation.
Whether working with human translators or machine translation models, providing relevant context on how the content is used is essential for achieving accurate and impactful translations. This could be previews of the text in situ in a document, or additional information on an industry specific term. Providing such context helps ensure that the intended meaning and nuance carry through to the target language.
One of the most striking paradoxes we encounter is when organisations invest significant time and resources in crafting their source content, only to expect instantaneous, high-quality translations. Whilst technological advances have streamlined many aspects of the translation process, achieving the same level of craft in translated content still requires careful attention and expertise.
However, this does not mean the process needs to be unnecessarily time-consuming. By implementing the right ingredients and processes, organisations can achieve both quality and efficiency.
At its core, precision in content localization is achieved through careful consideration at every stage of the workflow. It is important to also consider the interplay between the stages as that is what helps to elevate the precision of content, as each stage feeds the other with the required resources to support the delivery of quality.
Here are the key elements we typically think about:
Often overlooked in pipeline setup, this crucial first step involves deciding what content to include or exclude for translation, as well as identifying or generating additional context (e.g. fetching or generating previews, enriching content with definitions or annotations, etc). It also determines what adaptations can be made before or after translation. Automating these decisions ensures consistency at scale.
Depending on the capabilities of the platform used, this area can help enhance the process beyond just translating the immediate content, such as link localization, summarisation, content adaptation to multiple formats, and much more, driving significant value.
The difference between good and great translation often lies in the context provided to translators and AI models. This involves identifying and leveraging appropriate contextual information, either actively in producing translations when calling models, or as reference material such as previews or glossary definitions.
Building on the content processing capabilities, this enrichment need not only focus on the content being provided and can be extended to find relevant context information from additional sources.
The tools and interfaces used for translation, transcription, labelling, or writing significantly impact output quality. From basic features like concordance search to advanced AI-enabled quality checks, the right tools enhance what can be achieved at scale.
For many years features in such workbenches have been table-stakes, with tools often looking very similar, but with the integration of task-focused AI Agents, they can help enhance the productivity for those working on complex content.
With the proliferation of AI models in the market, choosing the right one for your specific language and domain is crucial. It’s easy to get lost in the hype of vendors claiming to have the “best-model” based on focused evaluation sets that don’t match your content.
Model performance remains dictated by what it’s seen in its training data, so it’s important to evaluate models and select the right one for your content. In practice, we find combining models with different focuses, such as translation and post-editing, with customisation for our client’s domain and handling of context can significantly boost quality.
Finally, the depth of domain knowledge in translation and review processes plays a crucial role in delivering precision content. Subject matter experts bring nuanced understanding of industry-specific terminology, conventions, and contextual implications that automated systems might miss - particularly given the limitations of the current generation of Quality Evaluation models. However, their value extends beyond individual translations.
By systematically capturing expert feedback and decisions through structured review processes, this knowledge can be encoded into the workflow through comprehensive glossaries, domain ontologies, and model tuning. This creates a virtuous cycle where expert insights continuously enhance the translation process, ensuring consistency and accuracy across all content while building a valuable knowledge base that improves future translations.
This introduction marks the beginning of our Precision Content Series, where we’ll explore each of the aforementioned Pillars of Precision Content in more detail. In our next instalment, we’ll deep dive into content processing and filtering – an often-overlooked area where significant value can be captured.
By understanding and implementing these precision-focused approaches, organisations can better navigate the complex landscape of content localization and achieve the quality their content deserves.
If you’d like to stay up to date on our latest blog posts, you can subscribe to our mailing list and/or follow us on LinkedIn.
