In 2016, one of the largest translation companies in the world announced it had been able to translate 20 times more content with Machine Translations than with human teams. Since then, Machine Translation has come a long way, with the rise of DeepL, a German-based company that raised the bar for machine translation by producing ever more natural-sounding translations at a sentence level.
Contrary to what many may think, this does not mean human translators will be fully replaced by machines. Instead, the role of translators is gradually incorporate tasks such as Machine Translation Post-Editing (MTPE). Read below to learn more about MTPE and how it can save companies time and money while providing new opportunities for translators.
Machine Translation (MT) consists of automatically translating a text in the source language into the target language using specific software. At a basic level, the software is able to recognize and replace a word in the source language for a corresponding word in the target language.
However, as technologies advance, companies are able to train machine translation software to leverage millions of human-translated segments of text to generate increasingly precise translations, including whole sentences. Those translations still need to be reviewed by humans, and that is where machine translation post-editing comes in.
Many language service providers (LSP) have been using machine translation as a way to increase the productivity of their suppliers, cut costs, and shorten turnaround times. With the growing demand for this technology, translators now need to include post-editing services in their portfolios to remain attractive to the market.
Instead of translating text from scratch, MTPE is based on editing an automatically translated text so that it sounds fluent in the target language.
Before starting a post-editing project, it is crucial to align expectations regarding the desired delivery quality. There are two types of post-editing services that can be provided, each one requiring a low or high level of effort from the translator. Check out the main features of these two processes:
Machine translation is undoubtedly faster than human translation. However, speed and volume should not be the only variables considered to measure the effectiveness of a translation.
As much as machine translation continually improves with the use of artificial intelligence, the role of the translator/proofreader/post-editor is key to ensuring that the target text makes sense to the reader and takes into account their cultural references.
Therefore, MT and human translators are codependent, since the algorithm cannot learn without the translator’s cultural point of view and linguistic knowledge.
When using MT to translate the source text, it is up to the translator to review the delivered content and compare it with the original material to identify issues. Some common issues are:
In addition, the MT should also be reviewed in terms of style to ensure that the text sounds fluent and natural in the target language. But sometimes, depending on the MT used or the context, you’ll see that not all machine translation segments are created equal. Some will require minor tweaks, some will be perfect as is, while others will require a full makeover.
But how do you know an MT segment needs editing or is good to go? One tip is using the two-second rule: if, after familiarizing yourself with both source and target text, you analyze the segment for 2 seconds and realize you cannot easily edit it, just discard it and translate from scratch, or edit a fuzzy match from the translation memory. After all, MTPE is all about making translators’ jobs easier, not harder.
Machine translation post-editing can be successfully used to improve efficiency, cut production costs and turnaround times in the following content types:
Satsuma has been working with Machine Translation Post-editing for over a decade. Our internal MTPE training course prepares our post-editors to be efficient without losing sight of quality. Our MTPE team is able to work with any engine or even indicate the best solution for each case. Talk to us if you are interested in improving your MTPE program or creating one from scratch.