The necessity of Pre-edit in machine translation tasks

The accuracy of machine translation is improving day by day, but its quality still requires human intervention. One of the main reasons for this is that machine translation can only process the original text on a sentence-by-sentence basis, and cannot take into account omitted subjects or objects, context, or nuances. In other words, even if there are words omitted in the original text for the sake of readability or conciseness, machine translation can only process the content written in the original text, resulting in a higher likelihood of unnatural or incorrect translations.

Therefore, in this blog, we will introduce methods to improve the efficiency of post-editing work through pre-editing (rewriting the original text to make it easier for machine translation) while giving examples of some content that has been omitted in the original text.

Table of Contents

1. Example 1: Cases where the subject is omitted
2. Example 2: Cases where the object is omitted
3. Example 3: Case where some content is omitted
4. Example 4: Cases where multiple contents are omitted

1. Example 1: Cases where the subject is omitted

●Original Text: Will be cleared by the following operation.
●Machine Translation: It will be cleared by the following operation.
　⇓
●Pre-edit: The value will be cleared by the following operation.
●Machine Translation: The value is cleared by the following operation.
　⇓
●After correction: The values are cleared by the following operations.

In this example, the subject "value" is omitted, so in machine translation, the subject becomes "It". Therefore, by adding "the value" to the original text and processing it again with machine translation, "The value" will be correctly reflected as the subject.

However, in most cases, only the subject is omitted, so in post-editing, it is only necessary to replace the pronoun, such as "It", with the appropriate subject. Therefore, pre-editing may not be necessary depending on the situation.

※As there are multiple values and operations in context, "value is" has been changed to "values are" and "operation" has been changed to "operations". Machine translation tends to have difficulty recognizing singular and plural forms of nouns in the original Japanese text, so caution is necessary.

2. Example 2: Cases where the object is omitted

●Original: Ability to change the display time for each user.
●Machine Translation: The time until display can be changed for each user.
　⇓
●Pre-edit: The ability to change the display time of the control bar for each user.
●Machine Translation: The time to display the control bar can be changed for each user.
　⇓
●Modified: The time for displaying the control bar can be changed for each user.

In this example, by changing "display until" to "show control bar" as the object of "time", which is omitted, the result of machine translation will be improved. There is a possibility that users may be confused if it is "The time until display".

※ If you want to emphasize "for each user", you can also use "For each user, the time..."

3. Example 3: Case where some content is omitted

●Original: Slide the DATA switch horizontally one at a time.
●Machine Translation: Slide the DATA switches horizontally one by one.
　⇓
●Pre-edit: Slide each DATA switch horizontally using a minus driver.
●Machine Translation: Slide the DATA switches horizontally one by one using a flat-blade screwdriver.
　⇓
●After correction: Slide the DATA switches one by one in a horizontal direction using a flathead screwdriver.

In this example, since "using a minus driver" is omitted due to being mentioned in the previous sentence, it is necessary to reflect its content in the original text.

In instruction manuals, it is important to pay attention as sometimes repeated content may be omitted.

4. Example 4: Cases where multiple contents are omitted

●Original: If not set, please have the teacher add and open from the file name editing, or have the student resubmit.
●Machine Translation: If it is not set, please add it from the file name edited by the teacher and open it, or have the student resubmit it.
　⇓
●Pre-edit: If the extension is not set, the teacher should add the extension from the field for editing the file name before opening the file, or have the student set the appropriate extension in the file name and resubmit.
●Machine Translation: If the extension is not set, the teacher should either add the extension from the Edit File Name field before opening the file, or have the student resubmit the file with the appropriate extension set to the file name.
　⇓
●After correction: If the extension is not set, the teacher should add it from the field in which to the edit file name before opening the file, or have the student resubmit the file with a proper extension set to the file name.

The original text here contains a lot of omitted information, so the result of machine translation is almost impossible to understand. In such cases, it is necessary to completely rewrite the original text according to the context and reprocess it with machine translation.

In addition, pre-edit can be done in real time on the machine translation screen, allowing for efficient work. Below is an example of pre-editing done on the screen for "4". (Tool used: our company's AI-powered automatic translation tool MTrans Team)

Machine translation is constantly evolving, but there are often quality issues due to various omissions in the source text (especially in Japanese). Therefore, it can be said that performing pre-editing and post-editing simultaneously in real time is an important aspect in utilizing machine translation in the future.

Author Information

Andy ParkMultilingual Translation Group
Japanese-English Translation Reviewer

・In my previous job, I worked as an IT engineer for about 4 years, and then I worked as an English conversation instructor for 8 years, where I was involved in developing educational programs and training instructors.
・Translation experience of 11 years, specializing in IT and business fields.
・Currently engaged in translation work and translation quality management, primarily focusing on FA-related products such as product manuals, help documents, and operation manuals.
- Responsible for evaluating and verifying the translation quality of machine translation engines.