Machine Translation and Post-edit in Medical and Pharmaceutical Translation

On May 27, 2020 (Wednesday), we held a joint seminar with Memsource titled "Shortening Lead Time for Medical Translation! Machine Translation/Memsource Utilization Seminar".

We introduced the use of machine translation and translation support tools as a means to achieve shortened delivery times for pharmaceutical companies and CROs involved in medical and pharmaceutical translation.

 

Introducing the cost reduction rate and post-edit for machine translation, which has received high praise.

 

Table of Contents

1. Cost Reduction Rate in Machine Translation

2. Quality Standards and Post-edit Level

3. Example of Post-editing in Medical and Pharmaceutical Translation

4. Documents Suitable for Machine Translation

 

 

1. Cost Reduction Rate in Machine Translation

The main reasons for introducing machine translation are "shortening delivery time" and "reducing costs".

By using machine translation and making no modifications, it is possible to achieve "shorter delivery times" and "cost reduction".

It is not possible to release or submit a document translated by machine translation to the market without checking or correcting it.

However, spending too much time on post-editing, which is correcting machine-translated words, will result in a paradox where the goals of "shortening delivery time" and "reducing costs" cannot be achieved.

 

One of the reasons why machine translation implementation does not proceed smoothly is the difficulty in establishing a workflow for post-editing, which involves making corrections after using machine translation.

However, by establishing quality standards and a machine translation workflow, it is possible to reduce the workload by up to half for certain translation items.

The following is a comparison table between the time spent post-editing after machine translation and human translation for protocols, ICF, and inquiry items.

By establishing quality standards and defining the overall machine translation workflow, it was found that we were able to reduce the workload by more than 20%, and nearly 50% in the case of ICF.

 

 

How about comparing time and cost in terms of cost and amount?

The following is a graph of cost reduction in a project that incurred a cost of 100 million yen due to human translation.

This is a successful case of cost reduction of 40 million yen. As the project was large-scale, there was an initial investment of 10 million yen, but it was found that there was a sufficient cost reduction effect even after deducting that amount.

 

 

So, specifically, how should we proceed with post-editing?

 

 

2. Quality Standards and Post-edit Level

In this article, we will define post-edit as "checking and correcting the translated text to ensure it meets the required quality by comparing it with the original text and the output of machine translation."

At Human Science, we have established 7 quality criteria items such as accuracy and fluency for post-editing.

Defines the items to be edited according to the target document, dividing them into "light edit" and "full edit" to determine the extent of editing.

Before starting the project, we will reach an agreement with the client on the desired level of quality and then begin work.

 

 

 

3. Example of Post-editing in Medical and Pharmaceutical Translation

The following is an example of a translated document in the medical and pharmaceutical translation field using post-editing.

 

 

Light edit is a correction from the perspective of "accuracy".

Correcting 2 mistranslations.

First, modify "on tissue of...a microscope" to "under a microscope on original tissue or..." due to the different modifier relationship.

The second point is a mistranslation of technical terms. "Signature" is incorrectly translated as "署名" (signature), so it is corrected to "marker".

Post-edit is guaranteed up to "accuracy", so corrections are only up to this point.

In full edit, the item of fluency is also added, and the text is adjusted to be more natural.

"Cancer is changing" is translated as "変化しています" in machine translation, but it is improving the quality by saying "変化しており・・・主流となった".

For other examples, they are available for download below. Please take a look if you are interested.

Machine Translation Translation Error and 5 Examples of Post-editing ~ Medical and Pharmaceutical Translation Edition

 

 

4. Documents Suitable for Machine Translation

Machine translation has been shown to have a significant impact on efficiency and cost reduction.

However, not all documents can be efficiently translated using machine translation.

There are areas of expertise with the MT engine, but below are listed items that are generally suitable for machine translation.

 

 

Documents with a strong internal tone are relatively suitable for machine translation. However, documents submitted to PMDA, FDA, EMA, etc. require accuracy, so it cannot be said that machine translation is a strong area. In the case of simple documents such as ICF, it can be said that they are exceptional for machine translation.

It is necessary to consider the type and purpose of the document, rather than simply lumping it together as machine translation.

 

 

We summarized the content of the seminar and introduced it.

In addition, we plan to hold the same web seminar in collaboration with Memsource in September 2020.

If you would like to receive seminar information emails, please register using the button below.