Due to the nature of the industry itself, regulatory compliance emerges as one of the biggest challenges, if not the biggest, regarding localization in the life-science industry. Regulation-related challenges come in many different forms: from adhering to tight time frames to ensuring that there is an audit trail for all translated content and complying with SOPs, life-science enterprises need to ensure that they meticulously follow every guideline and requirement set forth by regulatory bodies to maintain the integrity, accuracy, and safety of their products and services across global markets. Let’s break them down.
Life science companies face strict regulations in all their activities. Compliance is crucial, leading to numerous challenges, with few as significant as meeting stipulated time frames. For example, the new EU-CTR (European Union Clinical Trials Regulation) which came into force in January 2023, states that all clinical trial submissions must include documentation translated into all the languages of the trial participants. But the main challenge comes from having to respond to all requests for information in all of the languages within the space of 12 days – otherwise the application may be automatically withdrawn.
To be able to meet these requirements, life-science enterprises must be equipped with the right team of professionals to do it, but perhaps more crucially with the technology that enables them to carry out their work with as few delays as possible to meet the deadlines. The consequences of not meeting them can result in huge revenue losses for the company, as well as loss of competitive advantage and brand image in certain markets.
Delivering all multilingual content on time means nothing if the quality of the translation is not excellent. In other words, it’s not enough to just submit it through machine translation, but rather it must be post-edited by the necessary in-country reviewers (ICRs) and subject-matter experts (SME), to ensure that not only the terminology is accurate but that it is also understandable in the target language, making the process more of transcreation rather than simple translation.
For this, linguists and reviewers must be equipped with technology that enables them to perform these checks with all the information needed at hand, without incurring too many delays. The translation solutions chosen by life-science enterprises, therefore, must be equipped with robust terminology tools, customizable quality-assurance (QA) profiles, and reliable audit trails to track all changes made and ensure that there is full traceability of the process from start to finish to comply with internal SOPs and regulatory requirements. A wrong decimal here or there (e.g. a dosage of 1.32 instead of 13.2) or an incorrect translation or use of a term (sitotoxin vs cytotoxin, two terms that have nothing to do with each other) can have disastrous consequences.
Most life-science enterprises require the translation of millions of words every year into multiple languages, which only multiplies that initial number exponentially. Having this volume translated by humans within the stipulated time frames is unrealistic as it would be too costly and extremely time-consuming, which is why machine-translation engines become an essential tool in the localization process.
Being able to translate these high volumes therefore requires, once again, being equipped with the right technology that enables this to happen. When it comes to machine-translation engines, some are better trained than others in certain domains or languages, so having a deep understanding of which are better suited (which can be achieved by consulting internal or external localization experts) can save valuable time down the line by reducing the time of post-editing steps.
Automation is just as key. In workflows requiring multiple stakeholders’ involvement, offline communication and notifications, such as emails or sending files, are not an option when the time frames are so small. Therefore, automated workflows within a single solution are vital to move the project from one step to the next without incurring delay while also minimizing the risk of human error.
But what are the components of an efficient localization technology stack?
One of the important things to consider about localization within life-science companies is that they potentially have tens of thousands of employees worldwide, many of whom may be required to intervene at some stage in the translation process as requestors. But what happens if they are not familiar with the translation software? Each of them will use different tools or their own internal portals on a day-to-day basis, so to drive efficiency between teams, connectivity among the different tools becomes paramount.
You have Content Management Systems, translation management platforms, dependencies with other translation tools such as translation tools, QA Management tools or terminology databases such as MedDRA, and perhaps even proprietary tools like portals that enable inter-departmental communication, e.g., sending translation requests. For these reasons, a translation management system (TMS), the hub of all localization activity, must be able to provide strong APIs to all of the necessary systems required throughout the translation process.
The point of centralization is a key one. You need a translation tool like a TMS which can act as the neuralgic center of all activity. What this allows is for companies to have a single source of truth for all their content, paying particular attention to terminology as this is paramount in this industry. Removing the risk of linguists using different or outdated term bases, or being able to get queries answered in real-time as opposed to doing so via email, are huge contributing factors to on-time deliveries.
Also, as previously mentioned, life-science companies can have tens of thousands of employees, and training those who need to be involved in the translation process to use a TMS is unfeasible. Therefore, having everyone working on a tool they are familiar with that is successfully connected to the whole ecosystem also enhances efficiency considerably.
When choosing your TMS, giving your provider a comprehensive list of all the software tools your company uses for your global content and ensuring that they can all be interconnected via the TMS is a big step towards setting your program up for success.
When it comes to machine-translation engines, some are better trained than others in certain domains or languages, so having a deep understanding of which are better suited (which can be achieved by consulting internal or external localization experts) can save valuable time down the line.”
The first step in maximizing your translation efficiency is separating the different content types and understanding their priorities. For example, the level of validation required for the content of a clinical trial compared to internal communications is the complete opposite, therefore applying the same level of scrutiny to all translated content is simply not efficient.
This can be managed via a TMS by deploying different project templates for each content type and applying different workflow steps and tools per template depending on their nature. For example, you can have a specific template for content that does not require validation, such as internal communications that won’t be seen by the end-user customer. These templates should not necessarily include specialized stakeholders, such as SMEs or ICRs, as their time would be better spent working on valuable content. For sensitive content whose quality must be 100% accurate, you can create templates that include the SMEs and ICRs in the workflows. Creating different templates is not just a huge time-saver but also makes a big difference in allowing each team member to focus on the tasks that add value.
Another activity that can enhance efficiency is choosing the right machine-translation engine for each task. Not all machine-translation engines have the same capabilities: some are particularly useful for certain language pairings, and others for specific models or tones. Knowing the strengths and weaknesses of each machine-translation engine can drive efficiency by saving precious post-editing time during reviews. As mentioned above, you can find out the strengths and weaknesses of each via internal or external linguistic experts, or by asking your tech providers.
The third and last point is ensuring that non-linguistic stakeholders have it as easy as possible when accessing the localization tools they need to send translation requests to the right teams or perform reviews. It may be impractical to grant access to localization technology to thousands of users, never mind train them in it, so in this case connectivity becomes key. By being able to connect their original platform to the TMS, they can operate directly from it and send requests or comments from their original platform. For non-linguistic users who need to review, a TMS with a simplified view (i.e. without all the details like fuzzy matches, inline tags, or concordance searches filling up the screen) goes a long way in allowing them to focus solely on the content itself so that they can just read, edit, and submit.
Artificial intelligence is now the talk of the industry in localization, but the important thing to understand is that AI tools go far beyond ChatGPT. For many years now, AI-powered tools have been an integral part of localization processes, from neural machine translation (NMT) to proprietary tools of each tech provider, the quest for improved efficiency is nothing new.
One of the most useful AI tools for life-science companies is AI-powered machine translation and translation memories. These are particularly useful for companies with large volumes of content. Improvements in translation memories through AI technology enable the user to improve machine translation matches using high fuzzy matches coming from the translation memory. In layman’s terms, this means that thanks to AI, there is constant learning between the TM and the MT output. As a result, translations that required heavy post-editing and reviewing may no longer require as much effort, and projects can be completed in less time.
It’s not uncommon for life-science companies to have multiple vendors working on their multilingual content. In this case, it’s also not uncommon that each vendor used their own translation technology, which results in scattered language assets and, most importantly, a big risk in terms of consistency and accuracy. If you decide to unify all your language assets into one, bringing them all together is a daunting task in terms of the time it can take, an average of 50 hours for an average translation memory.
This is where AI comes into play in a way that most users are not aware of. TM aligners are AI-powered tools that align translation memories, detect any differences in the segmentation, and unify them automatically for consistency. The resulting output then becomes the basis of a translation memory. AI can also automatically insert HTML tags into target translations with a degree of accuracy close to 100%. These tasks may seem unimportant but we are talking of potentially hundreds of hours in time savings. All in all, when you think of AI in localization, don’t just think about GPT – expand your horizons and ask your tech provider what tools are available to understand how they can speed up your translation processes.
- Understand your industry’s challenges and the technology and processes that you need in order to overcome them successfully. Since regulations in life science are strict, the margin for error is much smaller.
- The proven path to localization success for life-science companies is to create an ecosystem of connected technologies that allow them to streamline their localization processes and involve everyone who needs to be involved without creating delays and bottlenecks.
- Managing translation quality is paramount, but it does not have to be a strenuous process. Equipped with the right technology, quality-assurance workflows can be seamless and simple for all stakeholders involved, even if they are not from a linguistic background. Prioritizing your content types for review and not using a one-size-fits-all approach is key to boosting time efficiency.
- AI technology has been in localization for a long time, long before the breakout of GPT. Therefore, it’s important to understand which AI-powered tools can help drive efficiency and how to best use them for your specific needs, as regulations in life science are much stricter than in other industries. And remember, the smart approach is not for AI to replace humans but rather to make them more productive.
With over 25 years in the localization industry, Iñaki Hernández-Lasa currently holds the position of Senior Implementation Specialist. During his career, he has specialized in different areas such as Enterprise Terminology Management, Linguistic Technologies, Machine Translation, Linguistic Engineering, and L10N platforms. He currently works on the implementation of language-technology software for enterprise life-science companies. Prior to starting in the localization industry Iñaki completed a research Masters’ degree in Translation Studies. He has been a guest lecturer at Lancaster University and Dublin City University and had different works published in The Irish Yearbook of Applied Linguistics.