15 Aug /19

Three Stumbling Blocks with Machine Translation – Particular Caution Is Advised for #3

Three Stumbling Blocks with Automated Translation – Particular Caution Is Advised for #3
Three Stumbling Blocks with Machine Translation – Particular Caution Is Advised for #3

Automated translation methods are improving all the time – in fact, it’s almost spooky how good they seem to be. This is thanks not least to the millions of users who feed online translation engines with vast amounts of their own data around the clock. It has to be fast and it has to be free. And these two criteria are often met – that is, if you define cash as the currency and not the information that is supplied for free in return.

This article covers three other weaknesses of machine translation. But one thing’s for sure: machine translations have nothing on the translation quality achieved by ‘real’ people (yet).

Three key stumbling blocks with translation engines:

#1 – Inability to read between the lines

A translation engine lacks any understanding of the text. While a human translator can skilfully pick up on and convey all the rhetorical subtleties and differences in meaning within a text, a translation engine ‘dumbly’ translates only what’s actually written. Ambiguous, woolly or overly wordy phrasing is often poorly or meaninglessly translated by machine translation systems.

#2 – Inconsistent terminology

There is no doubt that machine systems can translate – and thus apply specialist terminology – with increasing consistency and, thanks to the hard work of humans beforehand, ensure a consistent use of language. But total reliability is ensured only by a human translator, who can ensure consistency even if the source text itself is inconsistent.

#3 – Serious mistranslations

The biggest problem with neural machine translation (NMT) is that the text often sounds surprisingly good – and sometimes too good. You used to be able to spot a machine-translated sentence by the fact that, firstly, the wording was too clumsy or elliptical (i.e. incomplete) and, secondly, the text as a whole did not flow. Thanks to NMT, however, these problems have all but disappeared. Machine-translated texts nowadays can often read beautifully, and it is only on closer inspection or when the target text is compared against the source text that it becomes apparent that the target text is indeed too good to be true and has in fact strayed too far from the source text. The consequences of incorrect translations can be far-reaching – and, in the worst-case scenario, even put lives at risk.

Neural machine translation is not bad per se. It offers a range of compelling benefits and is ideal for very large or very urgent translations – or both. Compared with human translation, it is cheaper and allows companies that previously had no translation budget to finally have their documents translated. No longer are only the absolute essentials translated; now, the way is being paved for an integrated, multilingual communication and information strategy.

Are you interested in machine translation and require sensitive documents in multiple languages? Choose a professional language translation provider that uses its own closed translation systems to securely process your data and protect it against improper use. EVS Translations will be happy to advise you regarding your individual translation strategy, one that rules out neither machine nor conventional translation methods. For important documents, however, you are still well advised to rely on the tried-and-trusted craft of human translation so that you can impress the international market with authentic specialist translations.
You can reach us at :
+1 404-523-5560 (USA)
+44-115 96 44 288 (UK)
or use our contact form.