In the course of time, machines have taken over countless human tasks. A fully automated translation tool, such as Google Translate, would appear to be a logical step in this development. Unfortunately, opinions differ greatly on the added value of machine translations. On the one hand, enthusiasts swear by the significant time saving aspect and subsequently lower rates charged for machine translations. On the other hand, scrupulous translators are hesitant to believe that machines are currently able to deliver an acceptable quality.
Although translation tools already assist most translation agencies in their work, they still use human translators or editors. These supportive systems, better known as Computer Aided Translation tools (or CAT tools), check the text for any words or clauses which match previous translations. The system uses a translation memory where translations for each individual client are saved. The system suggests such matches which the translator then checks and includes in the translation.
Machine translations take this approach of CAT tools a step further. For instance, Google Translate and DeepL immediately show a result whenever a search term is entered. This technology entirely skips any human monitoring. Using a vast online database, the system computes the best possible match and then delivers a translation. This type of translation benefits greatly from the incredible speed by which results are compiled.
Tricky business
Nevertheless, this working method exposes a problem. To a great extent, the database quality and scope determine the quality of the machine translation. You will never draw up clean drinking water from contaminated wells. Similarly, machine translations based on polluted databases will only deliver poor translations. The machine lacks the skillset to assess the quality of the input.
Furthermore, machine translations fail to recognise the context of words and sentence structures. A translator understands from the context whether ‘crane’ refers to the machine or the animal. Ultimately, a machine makes nothing more than an educated guess at the correct meaning. Not to mention style conventions, consistency, imagery, and humor. Therefore, for the time being, translation is too complex a task to fully replace human translators.
The golden mean: post-editing
Clearly, these new translation technologies offer interesting possibilities. Translation agencies already experiment with post-editors: human translators who serve as an editor of the translation machines. The translation machine delivers an automatic translation of a particular text. Then a post-editor checks key elements that a translation machine fails to consider. For example, marketing texts often use humor, specific cultural references or a typical tone. The translator checks whether these elements are preserved in the target language.
For now, machine translations are mainly a solution for customers who have other priorities than quality. When a translation is urgently required, a machine translation with post-editing could be the perfect solution. Holiday websites already take full advantage of machine translations. Many accommodation reviews are often automatic translations which, in large measure, are grammatically incorrect. However, key words, such as ‘cheap’, ‘clean’, ‘beach’, or ‘good food’, are automatically translated. It still provides the reader with the most important information.
Regardless, a good translation remains a true craft.
Curious to know more about the possibilities of machine translations and post-editing services, and how Foxiz may help you with that? Don’t hesitate to contact us!