MT Post Editing: Best Practices for Machine Translation
7 September 2022

MT Post Editing: Best Practices for Machine Translation

Machine translation (MT) software has been developed and has come a long way since its conception, but nothing will still beat the accuracy that human translators will contribute to the fore. The problem with human translation is the slow turnaround time, especially for businesses concerned with a bulk of translation projects. 

A happy compromise to this dilemma is adapting machine translation post editing or MTPE. This will give you the best of both worlds regarding translation works which you will find out as you progress on the article.

What Is Machine Translation Post Editing?

Before we discuss what machine translation post editing is, let us look at what machine translation or MT is first.

Machine translation translates a text or script from one language to the next automatedly, without human intervention. Since the invention of artificial intelligence or AI, machine translation is constantly being improved and developed alongside it. 

Machine translation post editing or MT post editing, on the other hand, is the process where the translated text is edited, checked, and finalized by a human translator.

MT post editing concocts the perfect balance of AI speed and human accuracy and knowledge that each translated text should offer. 

There are many benefits to MT post editing, and the most integral of them is it saves money and time. MT post editing poses an increase of 350% in productivity.

Human translators can translate up to 2000 words on a good day, but this number can go up to 7000 words by utilizing the help of machine translation engines. 

Because it can be done in less time, MT post editing is cost-effective, and the quality is also not compromised.

Professional human translators do text or script checking, and some will even have a post-post-editing process. 

You might also like: How to Optimize YouTube Videos for SEO

Difference between Machine Translation Post Editing and Human Translation

There are many benefits to human translation, and one of them is the fluent localization that each finished text will represent, and each human translated work will embody. But, pivoting on the nature of the work or business, choosing the best tool that gives the most benefit will help the translators and business. 

MT Post Editing

MT Post Editing is Time-Saving 

One of the essential benefits of the combo of machine translation and post editing is time-saving.

Most translation engines have the ability to translate up to a thousand words in a matter of minutes, which is particularly beneficial for translation businesses that manage high-volume projects with strict deadlines. 

Since most translation engines will be cloud-based, many of their interfaces will be able to integrate different content management software that will be able to provide context for each word and sentence used in the text. 

Each translation work will retain its organization and true meaning even when translated into many other languages. 

Gone are the days when manual translations were done and professional translators were overworked. With MT post editing, time is saved since the translation is done automatedly, making the job easier and faster. 

MT Post Editing is Cost-Saving

Compared to human translations, machine translation coupled with human post editing work will cut the cost effectively.

Since professional translators are no longer involved with translation and are paid in localization mainly for their post editing works, people needing translation will spend less to have any text or script translated. 

The machine translation software will do the heavy lifting by providing a basic translation of the entire output.

Work for professional translators will involve ensuring that each output is true to the text’s original intent and localized to fit the region or language where it will be translated. 

The combo of a fast throughput and a selection of available languages available in various translation software will provide more rapid translation that will still ensure its quality since human translators will still make sure that they are true and close to the source text. 

MT Post Editing is Consistent and Accurate

One of the best things that only translation software will be able to provide is an intelligent translation memory and glossaries that translators or proofreaders can utilize and customize for their MT post editing processes. 

There are always regular subtitles for videos for TV, on the internet, or in book translations in most translation works. These tasks, which sometimes have industry-specific terms or repetitive dialogues, will be remembered by the translation software. 

Professional translators and proofreaders will then cut the time checking the source text, even for specific and already localized phrases, making for an accurate and consistent translated output each time. 

How to Produce High-Quality Machine Translation Post Editing Output

To make any MT post editing work successful, the first thing any translators or project managers need is to ensure that the MT output is high-quality.

If the translated output is excellent, the minor editing it will require leads to a faster turnaround of a translated project and more business for any translation business or team. 

MT Post Editing

Begin with a source text

Setting the MT translation for success will start from the very beginning, and that is with the source text.

Each translator or editor should make sure that the source text is pre-edited and created carefully so that the machine translation software and professional translators or editors are working with the best. 

When errors in the script are found earlier, there will be fewer mistakes down on the line. To ensure that the source is high quality, simply checking for any grammatical or spelling error in the text will suffice. 

Although many refinements have already been made in MT software, it will work better if the source text is as concise, accurate, and clear as possible. Twenty words are the ideal length of a sentence, and it should include correct and simple grammar. 

Choose the best translation engine

Translation works make resources reachable to anyone around the world. To make it easier for the translators and editors, choosing the best translation engines will make any MT post editing work more effortless and manageable for them. 

There is different translation software to choose from, and new translation engines are popping up and being developed with new features all the time.

Not all translation engines are equal, and some of the software will work better than the next, so choosing the best to do the translation effectively will save the team a lot of time and effort. 

There are several generic MT engines on the internet, and the best way to determine what works best is by testing and reading into each one based on the other translator’s experiences and your needs.

This might be the most time-consuming option, but knowing the pros and cons of each translation software will make MT post editing easier for everyone in the end. 

Different Post Editing Strategies 

Once your chosen MT engine finishes editing the source text and the translation, the post editing will commence. The amount of post editing to be done will depend on the quality of the source text and the MT engine, so it will differ from time to time. To ensure that every resource is utilized effectively, it’s essential to set expectations and strategies before post editing starts. 

MT Post Editing

Light Post Editing

The first mode of post editing is light post editing or LPE. By just doing light post editing to a raw machine translation, all the post editing will require are simple corrections to make sure that the translated output will convey the meaning of the source text. 

This does not mean that no editing will be done to the raw machine translations. An editor must ensure that the source document’s message and purpose are conveyed in the translated text. 

At the same time, to make sure that post editing is done fast and will lead to cost-effective results, editors and translators should aim to edit the machine translation as little as possible just as long as it is coherent in the language it is translated to. 

Full Post Editing

A more meticulous way to do post editing is with full post editing or FPE. With full post editing, the raw machine-translated text will be thoroughly reviewed and edited to make sure that there are virtually no errors in the full transcript.

If light post editing is simply focused on the essentials, full post editing will involve various factors. 

A factor involves the tone and style of the overall text document. It will also include an assessment that no grammatical errors in the translated language are made.

The translated words should be strictly localized for the language, from full-on sentences to idiomatic expressions. 

If a document requires a complete FPE treatment, translators and editors should ensure that each reader who will be accessing the translated texts should be convinced that what they are reading was written in their native language. 

This approach will require more time and cost than light post editing, but it will yield a higher quality output. 

Project-Specific Strategy

More often than not, requirements or guidelines will come with a translation project, and translators or editors need to glorify these requests by making sure that these jobs are tailored to the requirements. 

Think of the two post editing strategies as a spectrum with light post editing on one end and full post editing on the other and the meter of post-specific strategy adjusting between the two poles based on suggestions. 

Some clients will ask the translators to focus on a specific segment while doing LPE on the other parts of the source text, and translators need to be sure that they are saving time and money but still passing stellar work. 

Some work will also skip post editing entirely, and this can also be the case if the source text is meticulously checked before having the machine translation engine do its work. 

Post Editing Tools 

To save time and be more cost-efficient, there are a variety of software tools that translators and editors can utilize to make the job easier. Many machine translation engines can integrate various software to make post editing a breeze. 

The first tool to keep in mind is terminology management systems. This tool will guarantee the consistency that is needed for each translation output. It will include translation memories for repeated words and phrases and many other reference terms each translator or editor will need to fix the raw machine translations. 

QA tools are also there to be integrated into your translation engines. This tool will help the translators and editors doing post editing work see any issues in the source text as well as new errors in the translated script for MT post editing. 

Best Practices for Machine Translation Post Editing

If you want the post edited machine translation to be as high-quality as possible, the combination of machine translation efficiency and human work in post editing is the way to go. Aside from post editing, human intervention is needed in steps as early as reviewing the source text and others. 

Ensure that the source text is as accurate as possible

Preparing the source text is the first step to ensuring the best practice for MT post editing work.

Any errors like grammatical errors or longer sentences can cause complications to the machine translation in the end, so ensuring the source text is pre-edited before machine translation is essential. 

Some of the most common corrections in pre-translated works are changing the script to active voice, keeping sentences short, and using simple grammar that adheres to the subject, verb, and object structure.

If not edited on the get-go, one spelling mistake translated into five languages is five typo errors editors will need to correct in post editing. 

Many translation engines tend to be unable to translate sarcasm correctly, words with multiple meanings, and double-negative from the language of origin, so avoiding them or keeping them as literal as possible is important from the beginning. 

Pick the best Translation Management Software

The second point is making sure that you have the most appropriate and best translation software you can get for all your translation works. 

Some machine translation engines will be better for a particular language, while some will be best for a specific industry.

Define your translation niche and language, and filtering through different machine translation software will be easier. 

Other than the readily available translation engine on the interweb, one way to make the translation work more accurate in less time is by opting for a customized machine translation engine. 

With these, you will be able to integrate different apps and tools and customize the interface to use your data for easy localization, whether for the most used termbase or memories of previous translations. 

Set clear goals and guidelines for editing work

Setting clear translation guidelines and standards for MT post edited translation work will keep all output coming out of your office stationery and consistent.

This tip is vital for more extensive processes working with a team of translators and editors to bring out accurate and timely translation works. 

Are you opting to use oxford commas for every document? How many sentences should be in a single paragraph, or what voice should your work adopt to stay true to the outline? These are only some of the questions that need to be answered. 

There should be one place that each team member will look at to find all the information and guidelines they will need to keep all work consistent. 

Post Editing

Each translated work needs to be evaluated and continuously improved upon by skimming through post-edited results and client feedback.

One of the tools that translators and editors can use is the QA tools discussed above, which will help check any missed errors in MT post editing. 

With checking, editors and project managers can see where the works are lacking or where they can improve to keep translation works as time-saving and cost-efficient as possible. 

Translation engines won’t provide the data to tell you where to improve, so feedback from editors and translators should also be welcomed. Maybe it’s time to invest in more reference documents for easier localization, or source text should be pre-edited further. 

Final Thoughts

MT post editing poses the best of both worlds, and using this process for your approach will help you meet any client’s expectations and deadlines. 

The main goal of MT post editing is reducing a translation work’s turnaround time while still assuring that the work maximized quality by having human editors check and make adjustments to raw machine translations. 

It is continuously improving, so looking at the latest innovation to aid your team’s work will help tremendously and take feedback from the group and your many clients.