Part 1: Is the Introduction of Generative AI into Medical Translation Progressing? A Thorough Explanation of the Current Situation and Challenges

1. Can Generative AI Be Used in Medical Translation?</h>

In recent years, generative AI, represented by ChatGPT, has attracted global attention. While the use of generative AI is also expected in the field of translation, its introduction in the medical translation field, where expertise and accuracy are especially required, has only just begun. Many translators and companies are interested but have questions and concerns such as "How exactly should it be used?", "Is there a risk of information leakage?", "Will quality decline if entrusted to AI?", and "Will it really improve work efficiency?"

Related Blogs
>>Is it possible to improve medical translation efficiency with ChatGPT? Introducing real verification results using prompts!
>>ChatGPT and Medical Translation
>>ChatGPT and Medical Translation ② Are "prompts" effective in medical translation? We tried it

2. Current Status of Generative AI Implementation in Medical Translation</h>

2-1. Trial and Error in Utilizing Generative AI</h>

Currently, the use of generative AI in medical translation is at the trial stage. In some cases, efforts are being made to create draft translations using ChatGPT or to use it for checking translation results. For example, experiments are underway to automate correction tasks that previously required manual effort by linking chatbot APIs with translation software and refining prompts (instruction texts) to unify technical terms and adjust styles. ChatGPT has the flexibility to handle various translation tasks when given appropriate instructions, and it may be possible to obtain the necessary translation results even without large-scale pre-training data like traditional machine translation engines. For this reason, movements have begun to utilize generative AI in parts of the translation process, such as assisting in glossary creation and proofreading translated texts.

However, strict quality control and information security are indispensable in medical translation. Since highly confidential documents such as patient privacy and clinical trial information are handled, there are circumstances that prevent easy input into online AI. Even if it is convenient, one must be cautious about entering confidential documents into external AI services. Additionally, although the translations output by ChatGPT may appear fluent at first glance, there is a need for humans to verify risks such as mistranslation or omission of technical terms. In short, at this stage, it is not a situation where "using generative AI as is will solve all translation issues."

3. Challenges in Introducing Generative AI into Medical Translation</h>

The following concerns have been raised by medical translators and companies.

3-1. Security and Risk of Information Leakage</h>

"I want to try using ChatGPT for translation, but is it safe to input confidential data?" – Concerns about security and confidentiality obligations. There is particular resistance to entrusting business data to cloud-based AI services.

3-2. Lack of Know-How in Prompt Design</h>

"I don't know what instructions to give to generative AI to get the desired translation" – Lack of know-how in prompts (instruction sentences). It is still a trial-and-error stage on how to specify technical terms and writing style.

3-3. Verification of Expertise and Accuracy is Indispensable</h>

"Cannot fully trust the quality of AI translation" – ChatGPT generates fluent sentences, but there remain points that require human review, such as subtle nuance differences, factual inaccuracies, and formatting inconsistencies. Especially in medical documents, even small mistranslations can lead to significant differences in meaning, so the checking system cannot be omitted.

3-4. Will it really lead to increased efficiency?</h>

"There is doubt whether operational efficiency will improve even after implementation" – Since using AI involves learning costs and verification tasks, there are voices wanting to determine if it can truly improve overall efficiency.
Thus, while the improved performance of generative AI shows potential, there remain hurdles in terms of quality and security when directly applying it to medical translation. What becomes important is the assessment of "at which process and to what extent AI should be utilized." Careful verification is required to see how human expertise and AI strengths can be combined to achieve synergy.

4. Overview of Human Science's Initiatives Regarding Generative AI</h>

In light of these circumstances, Human Science, which provides pharmaceutical translation services, has launched an internal project to clarify the possibilities and challenges of utilizing generative AI. To address questions from the field, we are conducting verification using the following approaches.

4-1. ① Development of an In-House Tool to Utilize ChatGPT in the Translation Process</h>

We developed an AI-assisted tool in-house with consideration for usability and security so that translators can use it with confidence. This overcame issues such as "difficult to use" and "too risky to use," creating an environment where it is easier to try out on site.

4-2. ② Verification of Effectiveness through Demonstration Experiments Using the Tool</h>

Using actual medical documents, we examined how the introduction of generative AI affects translation quality and work efficiency. Specifically, we collected and analyzed data on issues detected by AI, its usefulness, and the impact on working time.

5. How to Leverage Generative AI in Medical Translation? Introducing Specific Initiatives from Next Time</h>

In this blog series, we will sequentially introduce the insights gained from this initiative. In the first installment, we will grasp the current situation and organize the challenges, >in the second installment, we will explain the content and verification results of the developed "ChatGPT Translation Proofreading Tool", and in the third installment, we will discuss the collaboration model between AI and translators and future prospects. Let’s think together about the future of medical translation in the era of generative AI.
Next time, we will take a closer look at the AI translation proofreading tool developed by Human Science and its demonstration results.

>Medical Translation Services
>Post-Editing Services

Free Download

Five Challenges and Solutions in Medical Translation

Introducing common challenges faced by many companies in medical translation and specific solutions.
Additionally, we present multiple success stories of cost and delivery time reduction.