Post editing of machine translation is an increasingly important part of translation agencies’ offering. Machine translation is here to stay, which means that machine translation post editing (MTPE) has become an essential service that many businesses rely on.
Today, I would like to explore MTPE in detail – what it is, why translators should care about it and how post editors and clients can work together to enhance the overall MTPE process. Shall we dive in?
Machine translation post-editing (or post editing or postediting, if you prefer) is when a human takes a machine translation (MT) and brings it up to scratch. MT post editing blends the skill and accuracy that the posteditor brings to the task with the speed and convenience of using machine translation.
Done well, MTPE can result in high-quality translations that save businesses time and money compared with translations that are carried out in the usual way.
I won’t delve too far into the history of machine translation here, other than to note that it has been around as a concept for centuries and as a reality since the 1950s (if you’re interested in translation history more generally, by the way, then check out the link below).
In simple terms, machine translation refers to the use of computer software to translate text (either written or spoken) from one language to another. In practice, creating a machine that translates with the nuance and finesse of a human translator has proven tricky.
Machine translation quality took a leap forward with the introduction of deep learning and neural networks from around 2014 onwards and it continues to improve today. It delivers impressive time savings, even with the need for MT post editing services factored in.
At Tomedes, our skilled posteditors work hard to enhance the quality of machine translations. To do so, they focus on tone, flow, clarity and accuracy, as well as on grammar. They add their nuanced translation expertise to the work that the machine has undertaken.
Tomedes’ MT post editors also take account of cultural sensitivities. Based around the globe, our post-editors don’t just deliver linguistic expertise but in-depth local knowledge. This is key to ensuring that translations don’t offend local sensitivities.
Appropriate technical knowledge is also a must. Tomedes pairs its MT post editing specialists with work that matches their experience, so that they can apply the right terminology (I’ll explore this in more detail below). It’s all about blending state-of-the-art tech with highly skilled human input – this is the key to successful MTPE.
Read more: A Very Brief History of Translation
Machine translation can deliver substantial cost savings to businesses, so many are embracing it as a core part of their translation workflow. This means that translators have a choice between sticking with their traditional methods of translation or adding a new service to their repertoire: post editing of machine translation.
When it comes to embracing MTPE, the key benefit for translators is that it opens the door to more work. That includes more work from direct clients and through translation agencies that are looking to recruit post editors.
Working with a document that has been produced by a machine, to produce a hybrid translation, is a different kind of work than translating a source document from scratch. It’s ideally suited to those who enjoy editing work as well as translation, as it blends the skills of both roles.
Translators who offer MT post editing services display to their clients that they are embracing technology and its benefits and that they are keeping one eye on the future of the language services industry.
Does your business need to connect with people (whether customers, investors, manufacturers, suppliers or anyone else) on a regular basis in other languages? If so, then it’s likely you’ve been tempted to use machine translation. It’s also likely that you have been disappointed with the results if you did not use a post editor.
If you have taken the decision to used post editing machine translation services, then there are steps that you can take to drive up the quality of the resulting translation. These include everything from carefully choosing which machine translation engine and post editor you use, to preparing your document for translation in the right way. I will run through some machine translation post editing best practise tips now, to ensure that you have every chance of success when it comes to your business translations.
Let’s start with how to choose your machine translation engine. This decision can make a major difference to the quality of the translations that you receive. Using a machine translation engine that delivers high quality results can lead to a significant reduction in the amount of time that your post-editor will need to work with your translation. As such, it can be worth looking at premium machine translation engines as well as free software, as the investment can pay off through a reduction in MT post editing hours.
The machine translation engine that is right for your business will also depend on which languages you need to translate. Some MT engines deliver better results with European languages, for example, while others specialise in Asian languages. It's a question of finding the right machine translation engine to meet your needs, just as you will need the right post editor.
As a general rule of thumb, neural machine translation engines will outperform statistical and rules-based translation engines, though results will vary depending on the content of your translation as well as the languages. DeepL, NMT Systran and Modernmt could all be of interest to you if you are working with languages such as English, French, German, Italian, Portuguese or Spanish. Tencent or Baidu might be a better choice, however, if you’re working with Chinese.
Regardless of which languages you need, be sure to research your machine translation options well. The tech side of the languages industry is improving fast, so there may be new entrants into the marketplace that can help you, or more established machine translation tools that have improved their abilities with different languages since I wrote this.
Just as you put care into choosing the best machine translation engine for your business, the quest to find the perfect MT post editor is also deserving of your time and attention. Several factors come into play here. First, clearly, you will need a translator who has experience of providing machine translation post editing. But that's not all...
Your post editor will also need to be a fluent speaker of both your source and target languages – preferably a native speaker of the target tongue. They will also need to have experience of working in your particular sector. Whether you manufacture medical devices, provide financial services or anything else, your translator will need to be familiar with the terminology that people in your industry and in your business frequently use.
You have a couple of options open to you here. You can work through a translation agency or seek to engage a posteditor directly. In either case, ensure that the individual who will be providing post editing machine translation services is suitably experienced. And if you plan to use a translation management system, now is the time to discuss that with your potential post editor, to ensure they are experienced/comfortable with it.
If you have chosen a neural machine translation tool, you should have the option to feed data into it before the translation work begins. This means you can furnish it with glossaries and translation memories that relate to the terminology used in your industry. Doing so should enhance the results of the translation process.
Likewise, it is helpful to furnish your post editor with all of the relevant linguistic information prior to them beginning their MT post editing work. Your post editor will also need to know about any elements of the document that you do not want translated: company strap lines, product names or anything else that you want to keep in the original language.
A discussion with your post-editor on how any queries should be handled, in addition to the above-mentioned machine translation post editing guidelines, is also useful at this stage, as it can save time once the postedit begins.
Another way to achieve successful results when using MTPE is to engage in a post of pre-editing. This is where you get your document ready for translation. Key to this is checking it for any errors. After all, the machine will not be able to pick up on them, it only has the chance to compound them and create confusion. The same goes for any content that is ambiguous.
It’s also well worth running some sample copy through your machine translation engine. Doing so will give you an idea of the kind of results you can achieve and whether the document you are working with is suited to this type of translation or not. I will expand on the kinds of documents that are not suitable for MTPE in a moment.
Finally, remember that the process does not end after the post editor finishes their first pass of the document. There will be an element of back and forth required after the MT post editing process takes place. A kind of ‘post postediting editing’ if you will.
This final element of the post editing translation process can reasonably be expected to take less time once the machine translation engine and editor are both more familiar with your terminology. However, it is always worth monitoring how long each element of the MTPE process takes, so that that you can assess its efficiency. For more on quality assessment of machine translation, you can click the link below.
Read more: How Is Machine Translation Quality Assessed?
Successful post editing of machine translation involves the translator working closely with the client. This relates to many of the points I've included above – terminology, queries and so on – and also to the kind of MTPE service being provided.
There is more than one way that post editors can approach this kind of hybrid translation work. Some of the most common MT post editing strategies that translators can use include:
· Light post-editing – this involves making as few edits as possible while still ensuring the clarity and legibility of the text. It’s a cost-effective approach to MT post editing that delivers speedy results.
· Full post-editing – as the name suggests, this approach involves a thorough post edit, with the post editor working on everything from style and tone to cultural sensitivities. This will result in a higher quality document than light post-editing, so is ideal for customer-facing documents.
· Project-specific post-editing – some translation projects will include a range of documents, some of which will require a heavier edit than others. In these instances, it can be helpful to devise a strategy that focuses post editing work appropriately on the documents that need the most attention.
Note that the kind of strategy the MT posteditor uses will impact how long the task takes to complete. This has clear implications for the client’s budget and so must be discussed with the client prior to the post editing taking place.
Just before I wrap this up, I wanted to add a quick word about when not to use post editing of machine translation. While MTPE is suitable for the majority of translations, if the best practice tips above regarding preparing documents and choosing the right MT engine are followed, it may not be best for more creative types of translation – or rather, for transcreation.
Transcreation is a creative type of translation that marketing translators may use when preparing documents for foreign audiences. Doing so can involve creative design work and a great deal of local adaptation, as well as translation. In these instances, it may be better to avoid using machine translation with a post edit and instead opt for using a marketing translation professional to undertake the entire job. As ever, it will depend on the document in question and the company it’s for – I’m just adding this in by way of a final thought.
MTPE can save businesses a great deal of time and money but it’s important to get it right. Hopefully the advice above should help with that. And if you’re inspired to learn more about machine translation and its role in helping businesses connect with the world, why not check out my thoughts on the top machine translation trends for 2022?
© Copyright 2007-2022 TOMEDES. All Rights Reserved.