Translation quality is usually hard to assess. How do you determine whether your translations are best represented by an automated translation service or a human translator?
I could list the benefits of a 100% human-generated translation, but you would much better judge the machine translation of the leading translation service DeepL yourself in an A-B comparison to the showcase translation of BlacKKite.
Machine Translation & Privacy Concerns
The fees for a machine translation service are charged as a fixed monthly rate (DeepL) or per translated character (Google & Microsoft). These rates obviously exclude the cost for editing. Compared to my translation rates, machine translations seem cheaper.
Free of charge translation is also available when you use a (limited) web service and are willing to surrender your data to help train and improve the translation algorithm. Your data is your currency.
The result of a machine translation service trained with a bilingual corpus continues to refer to that very corpus. One could argue that the translation computer gives a cleverly formed stupid answer to a complicated question. For translators, the answer is not necessarily black or white. They know when to avoid or rather emphasize gray areas.
I use terminology very consistently, for example, when machine parts in a maintenance manual refer to a drawing. In contrast, in a blog post about the differences between the translations of translation engines and human translators, I use a variety of synonyms for the term machine translation to keep readers engaged.
- Difference in format
- Related meaning
On average, 57.3% of the words in my Dutch showcase translation differ to a greater or lesser extent. This originality rate says little about the quality of the translation, of course, but it does demonstrate that buyers of translation services have a choice. Do they opt for commonly used phrasing or prefer a distinctive wording that contributes to a higher SEO score?
Purchasers of technical translations sometimes restrict the number of characters used for the translation relative to the original, e.g. when text becomes invisible if the character count exceeds a certain threshold. An English to Dutch translator needs to keep track of text expansion because the Dutch language often requires more characters for the same message than the English language.
Deleting is one of the translator's standard tools that is missing from the toolbox of the translation engine. Only a post-editor can truncate a computer translation. The word and character density of my Dutch showcase project are on average just a little below those of Philips' original manual. The automated translation tool uses on average ten plus Dutch characters more than the original manual.
Translation Turnaround Time
Machine translations excel in speed. The translation of your documentation will appear on screen in no time. With an average daily output of 2,500 words, I cannot match that, not even with the assistance of the most advanced CAT tool. However, keep in mind that a computer translation still needs to be edited by a human proofreader before the text is suitable for publication.
Moreover, speed is not necessarily something positive. A translation can benefit from the translator contemplating it for a while to further ponder the matter at hand. A fresh look can prevent mistakes and solve creative blockages!
For the time being, the human translator remains the leading force in high-quality tailor-made translations with respect for privacy.
When good is just good enough, the text needs to be translated quickly and cheaply, does not contain privacy-sensitive information, and the result will be post-edited by a proofreader, a machine translation could be a sound alternative.
Decide for yourself and order a Dutch translation now or request a free quote.