Post Editing Machine Translation: What You Need To Know  - Rubric (2024)

Looking to reduce translation costs for your global company? Machine translation can be a great way to do just that, by quickly and easily translating large volumes of text.

But have you considered post editing your machine translations?

It's important to remember that machine translation is never perfect. There will always be errors when a computer algorithm has generated the translations. That's where post editing comes in!

Post editing is when a translator manually reviews the translated text to identify and correct any errors. It's the only way to achieve a truly high-quality output with machine translation.

This can be an extremely time-consuming task. Sometimes the cost of post editing outweighs the benefits compared to manual translation.

Here's why post editing is a key step to ensure your translations are high quality…

What you need to know before adding machine translation

Global companies often think that machine translation (MT) will be a panacea. They'll think can just "throw" some machine translation at their content, assuming the technology will provide quick, usable translations at the touch of a button.

Yes, machine translation is a great tool. At Rubric, we believe that if you can use technology to speed up your process, use it.

But machine translation without strategy is a bad idea.

It can create quality issues and — most annoyingly for companies like yours — create more work than not using MT in the first place, unless you use it in the right way.

Post editing is the step that makes MT into a valuable tool.

What is post editing machine translation and how does it work?

Post editing is a process where human translators manually check a machine-generated translation for accuracy, making corrections as needed. It ensures that the translation meets the agreed quality standards for that piece of content.

Light post editing is when the human editor only edits the machine translation for minor errors, such as typos or missing words. The goal of light post editing is simply to make the translation understandable.

Full post editing is when the human editor performs heavy rewrites and edits on the machine translation, with a specific eye on the style of the piece. The goal of full post editing is to ensure that the translated content is stylistically appropriate to the specific needs of the content and your company.

The type of post editing you choose will determine how much work is involved. But either option can be a time-consuming task. You need to make sure you have the resources in place to do it effectively.

Also, what happens to your post-edited output? Is it dumped, or is it recycled for future use?

The benefits vs costs of post editing machine translation

The big question you need to ask yourself when considering machine translation is: Do the benefits outweigh the costs? Or does it make more sense to translate the content manually from the start?

The major benefit of post editing is that it improves the quality and readability of your MT content. This makes the content more understandable to your target audience and ensures that the translations reflect well on your company's global image.

The cost of post editing can vary depending on your needs and the type of post editing you require. Full post editing will cost more than light post editing.

The original quality of the machine translations will also affect the cost. If your English source content, for example, is of inferior quality, this will cause a low-quality machine translation. In turn, this increases the cost and complexity of the post editing.

Your cost-benefit analysis will include factors like budget constraints, deadlines, and specific language needs.

5 tips for using post editing machine translation successfully

1. Use domain-specific and language-specific data for your machine translation

Domain-specific knowledge is often essential for machine translation to be effective. When you're translating content for a specific industry or niche, you need to use an MT algorithm that "understands" the specifics of your field.

Using domain-specific data in your machine translation helps ensure the translations are accurate and relevant to your industry.

Language-specific MT is also an important factor. In our experience, some MT suppliers only work well for some languages. You'll need to pick the right supplier for your languages.

2. Improve the quality of your source content first

One of the most important steps when using machine translation is making sure that the quality of your source content is as good as possible. This means checking for errors, ensuring that all the relevant information is included, and making sure that the formatting is correct.

By taking the time to improve your source content, you can ensure that the translations are accurate. This reduces the required post editing work.

3. Use Terminology Management and Translation Memories for consistency and content reuse

You also want to ensure that your translations are consistent across all content. You can do this by using aTerminology Management database,which will help to ensure that all terms are translated in the same way. This is important for you to maintain a consistent brand image across all of your global communications.

The information stored in a terminology management database can also feed back into the machine translation, improving the quality of your translations. You can get your post editors involved in actively adding new terms and translations to the terminology database.

Similarly, a Translation Memory (TM) stores all previous translations, which aids reuse. When properly managed, this can reduce the cost of translations as you don't pay for the same translations twice.

Reusing post-edited output can be a good way to reduce the costs of future translation. However, you'll need to set clear processes for how you will you manage the differently translated content. The last thing you want is for all your translations to end up in the same "bucket," with a jumbled mess of human translations, raw MT outputs, light post edited content, and full post edited content.

One solution is to post edit all machine translations that end up in your TM. This ensures that you receive the most accurate translations first.

4. Run tests with your machine translation and provider first

It's important to run tests with your machine translation system and post editors before you put the content "into production," with help from your translation provider. This will help to ensure that the process creates translations that meet your required levels of accuracy and meet your business' needs.

By testing your processes in this way, you can also identify any potential problems that may need to be fixed.

5. Work for continuous improvement

As with all translation processes, high-quality post editing is not a case of "set it and forget it." You can and should try to improve your translation processes over time.

When your machine translation system makes errors — as it certainly will — use these errors as data to help improve your translation workflow. Get your post editors to give structured feedback on the common MT errors that you are seeing and actively improve the source content to reduce these errors.

Your first step for successful machine translation post editing

Machine translation can be a great way to reduce costs for your global company, but post editing is a vital step.

Now that you understand what post editing is and how it works, you can make better decisions about whether it's the right choice for your business.

With careful planning and the right translation provider, you can use post editing to successfully communicate with your customers around the world.