Machine Translation vs. Machine Translation Post-editing: Which One to Use and When?

To put it simply, machine translation has been used for many years to have a computer translate text from one language to another with no human intervention. Since the first appearance of AI (artificial intelligence), machine translation has been constantly developed and improved. Today, many MT engines can produce text that requires little to no editing. 

It’s important to say that even though MT is becoming more and more accurate, there is still not a machine translation engine that can produce better translations (and be used instead of) than human translators. Machine translation, however, can enhance and facilitate a translator’s work, and there is also a method of translation called machine translation post-editing built on MT. The point of this method is to pre-translate your text using a machine translation system. The target text is then edited by a human translator. 

Whether you want to go with plain machine translation or choose machine translation post-editing (MTPE) depends heavily on different aspects of your project. In this article, we’ll help you make an easier decision as to which type to use. 

Types of Machine Translation 

First, it's important to understand how machine translation works to be able to make an informed decision about whether to solely use MT in a project or to enhance it with human post-editing. Let’s look at each type of machine translation and see how each one works. 

Statistical Machine Translation (SMT) 

Statistical machine translation relies on bilingual corpora to produce translation. It uses these corpora (i.e., the source and the target of the same text) to come up with statistical analyses and to determine what would be the most accurate translation of a specific source text. 

The most obvious drawback to SMT is the fact that it does not take context into account. You can, however, still produce accurate translation output. 

Rule-Based Machine Translation (RBMT) 

Compared to SMT, rule-based machine translation relies on grammatical rules when it produces the translation output. It analyses the grammatical structures of both the source and the target language to create the translation. The sentences are usually grammatical and understandable, but from a post-editing point of view, they require close attention, proofreading, and editing by a human translator. 

Hybrid Machine Translation 

As the name suggests, hybrid machine translation uses a combination of the statistical and the rule-based types of MT. Although you get the best of both worlds regarding quality, hybrid machine translation still tends to require a large amount of human editing. 

Neural Machine Translation (NMT) 

Neural machine translation is the most advanced form of MT. It still improves every day and is used widely by translators, LSP, and by everyday people in their day-to-day lives. 

Neural machine translation teaches itself on how to translate by using a large neural network. This method is becoming more and more popular as it provides better results with language pairs. It is thus far the most advanced form of machine translation—yet it has its own pitfalls, especially mistranslations while the target text looks very natural - hence it's harder to spot issues when post-editing. Plus document-level context remains a challenge to all MT engines, for the time being (e.g. the Hungarian "ő" translated as "he" and "she" in different sentences). 

Machine Translation Post-Editing 

Machine translation post-editing (MTPE) is the process where a text that was previously pre-translated by an MT engine reaches its final form after being edited by a human translator. It is increasingly popular among translators and LSPs because it can strike the perfect balance between the speed of machine translation and the linguistic and material knowledge of a human translator. There are several advantages to this method compared to human translation. 

It saves you time 

Compared to human translation, MTPE can make you up to 350% more productive. According to oneword, a human translator can translate around 2000 words per day. By using the machine translation post-editing method, this number can go as high as 7000 words in a single day. 

It saves you money 

When you are able to perform more translations in less time, it is obviously more cost-effective. You can achieve the same quality with MTPE as human translation since the machine-translated text also receives human editing to make the text more readable and appropriate for the audience. 

Great for large amounts of text 

Sometimes there are projects where you have to deal with immense amounts of text to be translated within a short timeframe. Depending on the type of the text and the target audience, MTPE can be an optimal solution for such projects. 

To post-edit, or not to post-edit 

...Or in other words, when to solely use MT or when to post-edit your machine-translated text? How do you know how much post-editing is needed? There are times when you only need to know the gist of the text, so it may not be worth it to post-edit your translation output. In other cases, where readability is enough, only light post-editing suffices.  

To make your decision, you must ask yourself a couple of questions. 

Who is the intended audience? 

First, you have to consider your target audience. If you’re dealing with a text that is only going to be read by a small group (such as an intranet at a company), then it is perfectly acceptable to understand the context and the basics of the text. These might not even need any editing, but just some slight proofreading. 

However, when you are dealing with content that has a higher value and will be read by a wide audience, you’ll need to create more of an authentic translation output and the feeling that the text was indeed translated by a human. In that case, a more thorough post-editing plus proofreading is needed to create your target text. 

How accurate does it need to be? 

In other translation projects, different levels of accuracy may be needed—the amount of post-editing can also be determined by assessing this factor. 

Do you only need the basic words to understand the text? Then simply go with plain MT. Do you also need the output to be grammatically correct? Go with light editing. Will your text be visible to a wide audience and need to be more human-like? Is it highly specialized and it is very important for it to be perfectly accurate? Go for full post-editing. Of course, there are levels to MTPE, and the more you can assess the amount of editing needed, the better you’ll be at these kinds of projects as a translator. 

How to get started with machine translation 

If you work with machine-translated text and you understand how much post-editing is needed for your project, you can get straight to work in memoQ. Our TMS has a myriad of different MT engines available. Along with translation memories, memoQ can be a perfect tool for machine translation post-editing. 

Try it for free for 30 days!