Post-edit 4 Points | Human Science Co., Ltd.

As we enter June, half of 2018 has already passed. The highlight of the first half of this year would be the Pyeongchang Olympics. Many athletes have given us inspiration.
Did you know that there was this news at the Pyeongchang Olympics?

Norwegian Olympic team accidentally orders 15,000 eggs

According to reports, a Norwegian cooking team ordered 1,500 eggs using Google Translate, but ended up receiving 15,000 eggs instead. Fortunately, they were able to return the extra eggs and it became a funny story. However, if this were to happen in our own lives, it could have been a different story.

Mistakes in numbers can lead to major accidents. This includes not only amounts, but also time, speed, hard disk capacity, and so on. If you are conveying this information to a third party through a website or document, you must confirm that there are no errors in this information before conveying it.

With the emergence of neural machine translation, the accuracy of machine translation has improved dramatically. However, such numerical errors often occur. This is where post-editing comes in, which we will explain today.

'Post-edit' refers to the process of checking and revising the results (translated text) of machine translation. The person who performs this task is called a post-editor.
In the past, the general translation process involved the translator translating and the reviewer checking the translated text.
With machine translation, the translation is done by a machine and the post-editor checks and revises the original text and translated text. As a result, part of the process can be delegated to the machine, making it possible to shorten the overall translation time.

However, as mentioned earlier, there are some weak points in the latest machine translation technology. This time, we would like to introduce four points that we would like to correct during post-editing.

4 points to keep in mind during post-editing!

1. Number

2. Tags

3. Inconsistent Style and Wording

4. Terminology

1. Number

This is the result of translation using an online translation tool.
This is just a little game, so we won't look closely at the accuracy of the English text, but please pay attention to the red numbers. The translated text has different numbers from the original text.
The latest machine translation technology can generate very fluent translations, but very rarely, mistakes like this can occur.
The above example is an extreme case, but you definitely want to avoid having an estimate sent to your client with such a translation. Be sure to include a post-edit step to eliminate any mistakes in the numbers.

2. Tags

When translating tagged files such as HTML and XML, one must pay attention to the "tags".
There are various types of tags, such as formatting tags that modify text and variable tags that change the displayed content depending on the situation. When translating sentences that include these tags, it is possible for the necessary tags to be missing from the translated text or to be added incorrectly. Therefore, during post-editing, the structure of the tags is compared between the original and translated text, and they are rearranged correctly if necessary.
In addition, if a sentence contains many tags, the accuracy of machine translation can significantly decrease. In such cases, it is necessary to re-translate, making post-editing essential when machine translating tagged files such as HTML and XML.

3. Inconsistent Style and Wording

When translating, it is important to pay attention to maintaining consistency in wording and style so as not to confuse the reader.
In the example above, the structure of the English sentence is the same, but the structure of the Japanese sentence is different.
Also, common variations in spelling such as "server" and "server" or "user" and "user" can occur.
While it is not necessary to maintain such consistency in general documents, it is helpful to do so when translating procedures and manuals for the sake of the reader. This is also a point to keep in mind when post-editing.

4. Terminology

Finally, one thing to note is the "terminology".
The red underlined part above should refer to the same guide, but the translation (English) uses different terms. This kind of "inconsistent terminology" can lead to inconsistencies in the content and potentially result in a decrease in brand power. Let's make sure to properly adjust the terminology in the post-edit.

The following are the four key points to keep in mind during post-editing.
By making necessary adjustments to these points before sending emails to clients or publishing content, you can improve the accuracy of the translated text. Please give it a try.

At Human Science, we offer a cloud-based MT software "MTrans Team" that utilizes the latest machine translation technology for everyday business use, as well as "MTrans for Trados / Memsource", an NMT solution for professional translators.
These tools come equipped with terminology management and spelling variation reduction features, making post-editing less burdensome.
In addition, Human Science has released a new tool specialized in supporting post-editing, "MTrans Post-Edit Booster". We highly recommend this tool to those who are responsible for post-editing on a daily basis.
If you are interested, please feel free to contact us.

Machine Translation Inquiry