two men translating languages from laptop computer

Machine Translation Post-Editing (MTPE): Why You Need This QA Process

Machine translation (MT) software has come a long way since its inception, pioneered by SYSTRAN in 1968. You can now find advanced platforms able to efficiently and accurately translate critical corporate documents with incredible accuracy in seconds. These astounding machine translation output capabilities have transformed the way businesses approach language translation.

Table Of Contents:
What Is Machine Translation Post-Editing (MTPE)?
What Is The Difference Between MT and MTPE?
Who Can Perform MTPE?
What Is Human-in-the-Loop (HITL) Machine Translation?
What Is Expert-in-the-Loop Translation?
3 Reasons to Use Machine Translation Over Human Translation Alone
Best Practices to Follow for Machine Translation Post-Editing
The Critical Role of MTPE in the Future of Machine Translation
Experience the Solution: Test SYSTRAN’s MT Engine Today

However, even with a feature-rich MT platform, machine translation post-editing (MTPE) is still a necessity to achieve the best translation quality. You can think of MTPE as a hybrid approach between machine and human translation that offers the best of both worlds for translation projects. Studies have shown that professional translators exert less energy when editing machine-translated content than when starting a translation from scratch, which leads to numerous advantages when compared to raw machine translation.

In this article, we’ll give you a complete overview of why MTPE is so important for high quality translations and how you can implement it into your translation projects.

What Is Machine Translation Post-Editing (MTPE)?

MTPE is the process of a human editing a translation project generated by MT software, or post editing machine translation. The reason it’s so prevalent in any conversation about machine translations, regardless of the industry, is because it combines the efficiency of MT output with the nuances provided by human translation. Oftentimes, projects demand both touches, even if light post editing is the only thing necessary for quality results and fully accurate translation.

You'll begin by creating original content when you implement MTPE into your translation process. Then, you can check that content against your style guide. If everything is in order, that content can be submitted into your MT software before being professionally edited by a human translator to ensure it is grammatically correct. Professional translators will take the time to review the machine translation, translation memory, word and phrase matches, and other elements to ensure high-quality output. This is the process of training an MT translation software, critical when you are using AI-enabled machine translation engines.

To sum up the process in three simple steps, think of it like this:

Step 1: Your content is entered into the software, which proceeds to translate the source language into the target language.

Step 2: The MT software’s algorithm proofreads for spelling mistakes, grammatical errors, or other inconsistencies.

Step 3: Expert human translators review the raw MT output of the text and edit as needed for quality machine translation output. The translator will check for consistency across voice, tone, brand rules, preferred terminology, and more. They can use this part of the post edit process to begin the localization process (or build upon what’s been provided by the MT software translation memories).

The role of MT engines in the translation industry doesn’t as of yet cancel out the need for human post editing in some cases. However, it does redefine a human translator’s role in the equation. Translators no longer have to sit and translate a document word-by-word for hours upon hours for high quality results. That kind of manual labor eats away at precious time and resources. Instead, translators can save immense amounts of time by using fast, accurate MT software and concentrating their skills on polishing the final document with post editing.

What’s the Difference Between MT and MTPE?

Machine translation (MT) is one of the quickest, most cost-effective, and most valuable methods to create high quality translation. The best machine translation engines on the market can deliver fast results without compromising post editing quality in the slightest. MT delivers consistent output quality as well, and since one of the goals of effective corporate translation is consistency across specific language pairs, that’s a significant benefit.

Businesses in any industry can greatly benefit from leveraging the power of MT engines with post editing. You want to find an MT engine secure enough to minimize compliance issues, accurate and customizable enough to incorporate industry-specific terminology, and fast enough to maximize post edit productivity. While this may make it seem like MT software does it all, there is adequate space for machine translation post-editing effort that can improve the initial translation.

MTPE is a QA task ensuring your machine-translated content meets your brand standards. Depending on your needs, there are typically two main types of MTPE that may work for your next translation project.

Full post-editing (FPE) involves a thorough edit for all elements of brand voice, style, and consistency. It’s more in-depth and detailed than a less complex post edit, which is light post editing (LPE). LPE is the process of reading through the newly translated document and checking for glaring errors that may ha. It’s not as deep a process as FPE.

Both techniques have a time and place in any QA pipeline.

Who Can Perform MTPE?

mtpe-1Having native or expert language speakers available to perform MTPE is an ideal situation all companies should strive for. Localization efforts have so many advantages to an organization’s internal documentation, services, or products that the organization should invest in them. Native or expert language speakers are the individuals who can carry out those post editing efforts most effectively.

If you have these individuals on your internal post editing team, delegating MTPE responsibilities to those team members will facilitate the easiest transition. They know your brand’s style guide, are familiar with the organization, and can ensure consistent messaging.

Otherwise, you can hire freelance experts with your machine translation engine to conduct your MTPE processes. Be sure to provide them with a detailed translation style guide for your corporation to create the best results.

What Is Human-in-the-Loop (HITL) Machine Translation?

Human-in-the-loop (HITL) machine translation is another way of saying machine translation post editing. The term simply means that a professional human translator is integral to the process. HITL is when the power of machine translation is combined with the expertise and high-value offering that human intelligence provides in the post editing step. The humans involved in the process are critical to further developing the MT software and AI models that make machine translation so effective for high quality translations.

What Is Expert-in-the-Loop Translation?

The expert-in-the-loop translation method is closely related to HITL but is more specific to language experts modifying MT output rules based on previous output to improve future results.

3 Reasons to Use Machine Translation Over Human Translation Alone

If human editors still have to look over machine-translated content, what’s the point of MT software? Here are three reasons machine translation is a crucial addition to any organization-level translation process.

1. Speed Up Translation Times

Investing in a high-quality, secure, powerful machine translation engine takes resources. Still, the benefits of using MT software in your translation process are astronomical and makes post-editing QA tasks (such as MTPE) a breeze. Where your team would have otherwise spent many stressful hours manually translating a document to its target language, they can now use those hours to further innovate the post editing system as a whole.

2. Produce Higher Quality Output

Saving time means more than saving resources. When your human translators, whether in-house professionals or freelancers, aren’t spending all their energy on the word-by-word translation process, they have additional mental space to dedicate to adding increased value to the document with even light editing. They can edit the machine-translated content, localize it, dive deep into the voice and style to ensure all components are consistent, and more.

3. Increase Your Security Measures

Translation procedures and policies can be a confusing arena in terms of compliance laws. Many corporations don’t realize how crucial it is that all translation efforts be carefully compliant and have extensive security measures in place. Professional translators can still leave lots of things up in the air and increase the chance of human error.

Combining secure MT software with MTPE quality assurance means doubling down on security measures. This will help protect your organization in the long run.

What Are the Best Practices to Follow for Machine Translation Post-Editing?

To implement MTPE into your translation workflow, keep these best practices in mind for optimal results:

  • Keep your source text clean: Your machine-translated content can only be as good as the quality of the original text, so edit your source thoroughly.
  • Create a detailed multilingual corporate style guide: Since your organization uses multiple languages, your style guide must reflect their differences. Take the time to build out the guide to give your content a solid foundation to build upon.
  • Evaluate performance regularly: One of the best things about technology is that it’s always evolving. The participation of expert language speakers in the translation process should also mean continually interrogating the MT software’s rules and output to ensure quality.

The Critical Role of MTPE in the Future of Machine Translation

Machine translation (MT) isn’t a new invention. It’s been around since the mid-'90s when the first MT software was developed and to streamline translation processes. In essence, MT includes translation memories, and was designed to make translation faster and cheaper.

While MT software continues to be valuable to different industries, human translators remain relevant to the post editing process as key post editor resources.

As we mentioned earlier, this is what an expert-in-the-loop translation approach refers to for top notch translation services. The expert-in-the-loop translation method is where a native or expert language speaker is a post editor, and reviews machine-translated content and modifies the software's rules based on its output. Machine translation post editing completely changes the game, and because of neural machine translation features, this light post editing becomes even easier as machine translations improve.

This post editing method improves the output for future translations significantly. Refining the game-changing software solution is as essential to the process as the solution itself. Refining the solution enables the human translator to shift their focus from word-by-word translation and concentrate on innovation. This will lead to further advances in the human translation industry as a whole, making MTPE a critical piece of the journey overall.

Experience the Solution: Test SYSTRAN’s MT Engine Today

SYSTRAN’s neural machine translation (NMT) solution is one of the most feature-rich neural machine translation options on the market. It can provide astoundingly accurate translation quality, building on an extensive library of language pairs, dictionaries and industry-specific glossaries and supporting features like translation memory.

With over 50 years of history in the MT industry and a wide array of impressive features, SYSTRAN’s  is unlike any other, and is incredibly cost effective. You can experience this powerful post editing friendly machine translation solution for yourself. Test SYSTRAN’s MT engine today and find out more about how we can help you create better translations.

To learn more about specific translation terminology, check out SYSTRAN’s resource on every term you need to know.

Craig Wright