What is Machine Translation Post-Editing (MTPE)?

Machine Translation Post-Editing (MTPE) is a two-part process designed to save you time and money while you pursue your business goals in a foreign market. In the first part, the source content is translated by using an MT software. In the second, a human editor edits the translation generated by the MT software. The editor takes care to eliminate errors in the translation generated by the software and gives it the human touch it needs.

MT ("Machine Translation") is a process through which the content in one language is translated into another language using a computer software program.

No MT software today guarantees every single time that the generated translation will be 100% accurate and consistent, and flow naturally. MT is great at saving time but not completely reliable when it comes to quality. The machine is a machine after all: it will never match the human work quality.

Since the translation generated by MT software is never 100% accurate, consistent and natural in flow, you need a human professional translator-who acts as an editor-to check the translation against the source file and give it the final human touch it deserves.

The result? Translation that will help you engage with your clients or business partners in no time, and speak directly to their needs in a natural way.

Because the developments in AI technology finally meet the criteria for acceptable MT output (which still can't be used as it is: an editor is still needed to check it).

Suddenly MT has become the new buzzword across the world. The latest Neural Machine Translation technologies used among others-by giants like Google, Microsoft, Amazon, IBM, and Alibaba, and language tech companies like Systran, Lilt, DeepL to mention a few - are rapidly revolutionizing the translation industry.

Even regional tech companies such as Baidu in China and Yandex in Russia provide MT solutions.

Google Translate, which is an online MT platform, is used every day by millions of people around the world to translate billions of words into and from more than 110 languages.

This stuff is a bit technical. You don't need to know how machine translation technology works in order to use it! However, if you are fascinated by this phenomenal development in human communication, continue reading!

Machine Translation, also known as "Automatic Translation", is not a never-heard-of technology. It has been around since the days of World War II during which many MT tech trends emerged. We have seen primarily three trends or types of systems:

• Earlier pieces of MT software, called "Rule-Based Machine Translation" tools, operated based on grammar rules. Grammar rules of languages were incorporated into a piece of software fit with a dictionary. This system was a good start but the output hardly made the source text understandable. Imagine this: You couldn't even get the gist of French news if you were to use this piece of software!
• Then we saw the emergence of Statistical Machine Translation tools. Translations are generated by using statistical models to identify possible matches of words and phrases between two given languages. These tools leveraged on large, bilingual text corpora, which contained translations that had been previously done by human translators. The accuracy of statistical MT tools was better than that of rule-based models but not fully usable.
• Then came Neural Machine Translation (NMT) systems . "Neural" because NMT systems work like the neurons in our brain. Instead of analyzing words and phrases separately, NMT systems process whole sentences-from the first to the last letter- just like human brains do.
• Unlike previous MT systems, NMT systems are a marvel of technology because they have the ability to ‘learn' and ‘get trained'. NMT systems run learning algorithms developed by AI experts. They are trained using huge parallel corpora that contain previously translated content in a particular language pair.
• When fresh content that needs to be translated is fed into the NTM systems-also called "engines"-the NTM systems translate such content using the artificial knowledge they have already acquired through training.
• Previous tools analyzed sentences partially and could generate translations only at word or phrase level. NMT systems, on the other hand, succeeded in grasping whole sentences as complete translation units.
• This means that these systems generate more accurate, naturally sounding translations that better capture the gist of written text. In some cases, we obtain over 90+ percent accuracy.

NMT systems are almost a miracle-a dream come true for the humankind. NMT already demolished the language barrier for millions of people. Very soon, people will talk to each other in different languages using some small gadget that captures the speaker's voice and translates instantly. Machine Translation is here to stay.

By adding human editing to this quick, highly efficient process, we make translations more useful and reduce the translation cost by a huge margin.

Yes, you can (unless your hands are tied). But should you? Probably not. Why? Because it can be risky.

Here are some types of texts you shouldn't translate using a machine:

• Creative texts-used in culturally specific marketing environments or fiction-are not suitable for MT. Yes, the machines are "mechanical": they can't feel the text and emotions.
• Legally sensitive texts should not be machine-translated. If your translation is meant to be used in a legal setting (say, in a court or police station), please check with your local authorities to make sure MTPE is accepted in your jurisdiction.
• Medical records should not be translated using an MT software. MT is not always accurate, and this could have negative consequences in a patient's life.

Ailaysa always has your best interests in its mind. For this reason, whenever you feel that MT will do more harm than good for you, we will react immediately. You will receive an email that reads something like this: "Hi, your content is not acceptable for MT. We found lots of errors in the translation generated by the MT software; the translation should be done by a human translator". If you wish, we can give you some samples of the output for you to judge for yourself.

Sounds too risky? No problem! Just send us an email and we will provide the best solution for your needs.

Even after many decades of research and development, MT is still an emerging technology.

For this reason, many languages are still either not supported at all or under-supported. The good news, though, is that MT language list is growing every year.

The MT accuracy level for a given language pair is one of the factors that determine the cost. The more accurate the MT is, the quicker the editing process will be.

Ailaysa uses many state-of-the-art MT engines developed by various companies. Ailaysa doesn't develop any AI engine. It will simply use the tools that best meet your needs.

The type of MT software to be used will depend on language pairs and subject matter. The process may change depending on what you are looking for.

Our editors, who provide post-editing services, are professional translators who work in the language pair you need your texts translated into or from, and specialize in editing MT output (translation generated by MT software).

Even the most sophisticated MT engine can't generate 100% accurate translations for you. And that's where our editors come into play. They use their skills, expertise, and experience to:

• Eliminate grammatical and syntactic errors
• Guarantee that no sentence, phrase or piece of information is lost in translation
• Enhance the style so that your message speaks to the heart of your reader/audience
• Make sure that the edited text is culturally and legally appropriate
• Make sure the correct terminology is used consistently throughout the text
• Use proper spelling, style and punctuation.

It depends on what you need. What we can say with certainty is that it's cheaper than human translation and human editing services provided by thousands of translation agencies around the world.