20/08/2024

The latest print edition of the trade magazine Healthcare Marketing includes an article that looks at how artificial intelligence in medical translation is changing the industry. The magazine interviewed four translation service providers about the opportunities and challenges presented by using this technology, particularly in the context of the high standards required in professional medical communications. The experts included Jasmin Nesbigall, Head of MTPE and Terminology Management at oneword. Here we summarise the most important insights from the article.

Healthcare Marketing describes the current situation very aptly in the article’s introduction: “AI-supported translations have been an everyday part of life for a long time. They can make the translation process more efficient and cost-effective. They can also be a game changer in specialised medical translation in terms of productivity and cost savings. However, the process requires a high level of quality assurance and an understanding of regulatory aspects and terminology units.” Jasmin Nesbigall talks to the trade magazine about what exactly this requirement looks like and how the technology proves itself in everyday practical use.

Healthcare Marketing (HCM): AI development is progressing at a rapid pace. What is the general impact of this technology on the structure of everyday life at agencies and, as a result, on the world of translators?

Jasmin Nesbigall (JN): The translation industry experienced the hype that ChatGPT triggered at the end of 2022 in a similar way back in 2016 with the breakthrough of neural machine translation. Large language models, such as GPT, are multilingual, so they can also provide translations, even though they have not been explicitly trained to do so. However, other language-related usage scenarios, such as checking and correcting source texts, terminology work or adapting a translation to specific stylistic or formal requirements, are more exciting for our industry. AI can definitely support many process steps in these scenarios.

HCM: What can AI translators achieve in the medical translation field for Rx, OTC and medical technology and what are their limits? (Note from oneword: Rx: prescription drugs; OTC: over the counter, non-prescription medicines)

JN: Machine translation can also be a real game changer in specialised medical translation in terms of productivity and cost savings. However, it is always important to consider the risks: no MT or AI system delivers error-free translations. The risk arising from translation errors can be particularly high in medical translations. AI systems also involve inaccuracies and hallucinations. Therefore, the human mind is a crucial component when using artificial intelligence. AI can be a very effective aid to the process, but it should never take sole control. It’s important to have well thought-out processes and to use the technology in a targeted way in line with the company’s requirements. It’s often worthwhile to seek an expert outside perspective and advice to assess which processes would benefit from AI.

HCM: What are the quality assurance challenges in AI-based specialised translation?

JN: The errors resulting from MT and AI are sometimes very subtle and are hidden under a cloak of linguistic elegance. This is because both NMT systems and large language models formulate sentences very eloquently. It is therefore important to check the translation output carefully and compare the source text with the target text. This means that human checks are crucial, because AI systems create texts based on word-level probabilities. In other words, they correlate words based on how often they occur together. This can lead to omissions or very free additions when the text is generated.

HCM: Will technical developments change the language service provider’s business model in the medium to long term?

JN: The language industry is a very dynamic sector that is constantly adapting to new technologies. LLMs are a technology that can provide useful support in many parts of the translation process, both in routine tasks, such as project creation and file preparation, and in tasks that used to be time-consuming, such as compiling and analysing data. This requires integration into established tool landscapes and must enable human interaction. The output from AI systems, for example on quality assurance, rewording or terminology work, must be traceable and verifiable so that experts from translation and project management can intervene at any time.

The interview was published as part of the article “Medical Translation: Translators focus on post-editing and quality assurance” in the July 2024 print edition of Healthcare Marketing. The issue is available by subscription or as a single order from the publisher.

Do you need professional, secure and high-quality translations in medical or medical-technical communication? Then talk to us. Our medical translation experts will be happy to advise you on the processes and options available.