If texts need to be accurately translated into another language in a legally certain manner that complies with data protection requirements, pure machine translations are out of the question due to various drawbacks.
AI and machine translation
AI is gradually dominating our lives. Chatbots such as ChatGPT can even generate texts in foreign languages. However, with their output based on probabilities and their lack of understanding of texts, they are not the number one choice for translations.
Machine translation tools such as Google Translate, which is integrated into the eponymous leading search engine, are more suited to this.
However, machine translations still have significant drawbacks that cause problems in professional environments. These particularly include the following:
Incorrect or unsuitable terms are often used due to the context being disregarded. This especially applies in the case of stand-alone terms that do not form part of a sentence.
Terminology is not used consistently. Terms that frequently appear in a text are often translated in different ways. In contracts, this can sometimes even result in inaccurate content.
Set phrases, word plays, puns and sayings are not detected. This results in an incomprehensible or inaccurate text in the target language.
The style is poor. The more creative the text to be translated, the worse the problem.
Even with frequently used translation programs, technical constraints make it impossible to ensure that the source and target text completely match, meaning that errors can occur. The risk of errors is particularly high when it comes to legal texts, for example, which often contain very long sentences with multiple clauses.
There is often a lack of differentiation between country-specific variants (e.g. Portuguese for Portugal and Portuguese for Brazil).
No localisation occurs. No consideration is given to the standard date, time, currency and temperature formats used in the target country. The same applies with regard to units of measurement and conversion values.
Reference materials, existing glossaries and other terminology specifications cannot be considered.
Proper names are often translated as they are not identified as such. Conversely, terms are sometimes incorrectly regarded as proper names and therefore not translated even though they should be.
It often causes problems if multiple languages appear in a single text, for example due to somebody being quoted in a different language.
It is not usually possible to create terminology databases or integrate them into the translation process.
Data protection issues arise if an online machine translation tool is used and the text to be translated contains sensitive or confidential data.
Finally: machine translation tools do not accept responsibility for errors and their consequences.
So, are machine translations completely unusable? No. When it comes to general correspondence or garnering a basic understanding of a text, Google Translate and co. are extremely useful and a good choice.
In the case of professional texts, texts intended for publication, texts where the right terminology matters or texts to be used as legal documents, on the other hand, pure machine translations should not be used.
Depending on the language combination, specialist area and type of text, it can make sense for texts to be machine translated and then edited by a professional translator. This is known as machine translation post editing and can involve ‘light post editing’ or ‘full post editing’. Whereas the aim of ‘light post editing’ is to create a ‘readable’ text, the aim of ‘full post editing’ is to create a text with the same high quality as a human translation.
Alpha Communication does not offer ‘light post editing’ as this does not live up to our quality standards.
‘Full post editing’ is considered in cases where the language combination, specialist area and type of text make it feasible for the post editing of a text to involve less work than a conventional translation but still offer an end result of the same high quality.
We will gladly assess the feasibility of this service for you on a project-by-project basis.
Even if the translation process involves machine translation, we are naturally still able to integrate the terminology databases that we use for all our clients. This ensures that their company-specific terminology is used consistently and results in significant medium and long-term cost savings.
For more information about this topic, please do not hesitate to get in touch.