We all know that machine translation (MT) services can be used in some contexts as a tool to increase productivity in the translation process. The availability of well-known APIs for web services has made this method very popular. However, it means that users of these services demand a faster turnaround. They want them to be customized, within their area of expertise.
Ever since the days of rule-based and statistical machine translation, technology has not stopped evolving. The decision between using machine translation or human translation at the outset of a localization project is becoming less relevant because of the widespread use of online translation tools like Microsoft’s Bing Translator, Google Translate, or DeepL.
More and more businesses, language service providers (LSPs), and translators are recognizing the benefits of machine translation post-editing (MTPE), which involves human linguists editing machine-translated content. However, these tools cannot be easily customized for a particular use case, for a particular client or application – or they simply cannot be customized at all.
MTPE (Machine Translation Post-Editing) or PEMT (Post-Editing of Machine Translation) is now largely considered a very viable and affordable alternative to translating text from scratch. It saves time and money, while still producing high-quality translations, and most importantly, it is scalable because AI only needs more computing power to produce more content.
When deciding on the appropriate translation method, it is helpful to consider the content type and language pair.
Generally speaking, machine translation is perfectly suited for predictable, structured content, such as technical, legal, and intellectual property documentation, even healthcare translation services, as well as internal communications.
MT can struggle with uncontrolled, unpredictable language where creativity is needed: marketing materials and other customer-facing content may be machine-translated as an initial step, but nobody would publish public-facing material, slogans, or beautiful product descriptions without a more human touch.
You may be interested:
Speed: Post-editing is without a doubt much faster than traditional human translation, as the machine translation engine does the bulk of the work in seconds or minutes – or instantly in the case of live translation of websites or user comments.
This can be a major advantage for an enterprise that needs to translate large volumes of content quickly (or instantly).
Accuracy: Post-editing is just as accurate or even more than traditional human translation, as human linguists correct any errors that the machine translation engine makes. Well-customized translation models (engines) will contain the client’s specific terminology within the training set and scientists will have trained the model with expressions and style from the client.
MT output avoids mistranslations that can happen as a result of accepting a fuzzy match coming from translation memory, for example.
Cost-effectiveness: Because of the improvement in turnaround times, hands required, and file processing, post-editing is more cost-effective than traditional human translation, as the human translators only need to edit the machine translation, rather than think and translate the text from scratch.
Overall, MTPE is a viable alternative to traditional human translation for a wide range of content types. It can save businesses time, money, and resources, while still producing high-quality translations.
We employ the term “raw” when machine translation is published or used without human revision (no post-editing). As a rule, we recommend using raw machine translation only in the following cases:
Low-visibility content like internal documentation;
Understanding content from other companies (their websites, their social media posts for sentiment analysis, etc.) in what we call gisting to obtain an idea of what is being said;
Foreign language bids and tenders (again, to ingest knowledge by gisting);
User-generated content like online service or product reviews, for which consumers generally have grown accustomed to quick online machine translation;
Content that will not affect your brand’s image;
Ephemeral content (low shelf-value) texts like chatting to clients or email support messages, customer inquiries, etc.
Large bulks of content with a short turnaround, such as hundreds of product descriptions that need to go live quickly or feature and information updates for which you have a customized translation model;
As MT developers, here at Pangeanic, we recommend the use of raw machine translation only in these situations and always after understanding that your developer is building in-domain translation models with your own material and terminology or farming bilingual documents or texts that are highly relevant to your domain or industry.
In the times of statistical machine translation, MTPE used to be divided into either light post-editing (LPE) or full post-editing (FPE). This no longer happens, as neural machine translation developers are constantly acquiring Data-for-AI in order to improve their models and update them with fresh content.
Large language models (LLMs) like GPTs have also shown amazing translation capabilities in the top ten world languages, but less so in low-resource languages – that is, languages with fewer speakers where fewer resources are available. Nowadays, all post-editing work is fairly light and consists of “humanizing” the version created by the translation model, verifying no obvious machine expressions remain, checking terminology, etc.
One of the best combinations is called TM+MT, which means that your translation company will put in place a system whereby each sentence is checked against a translation memory (TM) so that if it has been previously translated, the previously translated material will be leveraged from it. The new content will be sent to the custom MT system and content that has been slightly modified is sent to the translator (say the color of an automobile in the case of automotive translations has changed from white to red).
Modern translation companies, however, are coming to the conclusion that it is often not worth applying the three methodologies and if any part of a sentence has been modified, it is sent to the MT system in order to shorten production times.
MTPE has many advantages and guarantees that the brand style is adhered to in other languages.
As a general guideline, the below are perfect use cases for MTPE:
Technical documentation, user manuals, and service manuals, as they need to be 100% accurate both for users and technicians repairing or servicing your brand’s products;
Blog content and sections of your website;
Some SEO content like alt text in images that is not seen by users, only by search engines, and needs to be SEO optimized;
Product descriptions so they clearly state the product’s features or benefits without any room for ambiguity.
Content of medium visibility or relative traffic that needs to be as accurate as possible: knowledge bases/knowledge centers, alerts, etc.
Recommended Reading:
It is generally advisable to avoid machine translation when the goal is to engage, entertain, or reassure the audience or when cultural nuances can affect the message. When it comes to content that can affect the way your brand is perceived, high-traffic pages on your website from which users don’t often bounce, human-to-human communication, or content/documentation with a long shelf value, human translation services are the recommended option.
A human “finish” is best, meaning a human translator may need to recreate the message in the target language in a non-literal way—a process often known as “transcreation.” This is the case for top-quality client-facing content:
Brochures/flyers and all types of printed promotional material
Home pages
Online advertising
SEO content (often adaptation is required rather than a translation)
Relevant web content
Top blog posts
Newsletters
Press releases
Top product descriptions
Thanks to our PECAT online tool and our document translation services using ECO, you do not need to worry about the choice. You just need to consider the privacy of your content. If you are happy sending your content to a free online engine, just click a button to select the engine. We strongly recommend, however, that you use your own translation engines (translation models) as developed by your MT services provider and that all your relevant processing takes place within the privacy of a platform governed by a contract.
Beware of translation management systems (TMSs) choosing the “best engine” for you after running an analysis of your files or content to be translated: you are sending and sharing one of your most precious assets, your own content, with free online engines across the Internet. Until not long ago, and even nowadays in many cases, the price paid for such free services is handing over your post-edited material to such engines.
Centralizing all those decisions in a translation management system with privacy in mind creates a “source of truth” for all translation activities (translation memories, machine translation, post-editing, and re-creation of the source file) and helps you to successfully deploy 3 features in your content translation journey:
Optimal custom MT engines for your content type or general online engines for low-value content;
Built-in quality estimation for an improved post-editing experience and productivity;
Automatic metrics to stay on top of productivity, turnaround times, and cost savings.
Different translation projects may require different levels and types of sophistication. This is one of the reasons why utilizing MT as part of a TMS is so beneficial.
To get the best out of MT, you need to be able to assign the right engine for each type of content—and the most robust translation management systems have plugins or application programming interfaces (API) that connect them to different MT engines.
The most advanced systems even offer the ability to automate the selection process based on artificial intelligence or algorithms that scan the content and match it to the optimal MT engine.
An essential part of successful MT implementation is knowing where to direct post-editing efforts.
When you are able to measure the quality of MT output automatically, you can focus on the important segments instead of wasting time and resources where raw output is already of good quality.
This removes the guesswork from MT and improves post-editing efficiency, and it constitutes yet another reason for using MT integrated into a TMS: the most sophisticated systems include automatic machine translation quality estimation capabilities that can identify which segments need more attention than others.
As mentioned above,
MT is attractive to many organizations because it offers the potential for increased productivity, faster turnaround times, and ultimately, cost savings.
Not all MT engines deliver on this promise equally, so having the means to compare how different engines can affect the process is key.
A strong TMS lets you track the time and expenses of any translation project where MT is applied. With multiple MT engines in use, these metrics can be a strong indicator of an engine’s value: Is it increasing or slowing down the translator's productivity? Does it indicate improved efficiencies over other engines over time? The answers to these questions will give you a better sense of its capabilities.
Our ECO platform allows you to use our machine translation (PangeaMT) and to create your own private and secure translation ecosystem with Deep Adaptive MT engines. It is also trained to improve the AI system according to your company's needs.
Needs documents in other languages
Translates large volumes of text
Needs to cut costs and improve productivity
Is in the process of internationalization and needs documents in different languages
Takes data privacy seriously
Fully integrated into ECO, the advanced machine translation (MT) features offered by PangeaMT allow you to:
Start translating immediately and efficiently without the need for developer time or effort, using fully managed and customizable MT engines.
Manually add any generic or custom engines.
Enjoy unlimited machine translation for post-editing workflows, allowing linguists to work more efficiently.
Translate all kinds of documents.
Work with the best engine, automatically selected according to the language pair and the type of content depending on the sector.
Integrate with popular APIs.
Ensure that MT engines use your preferred terminology with the correct morphological inflection, reducing post-editing efforts.
Save time and money.
Translate in over 500 language pairs
Work on any type of content in a secure and private environment.