If a layperson only roughly understands a few words or passages of a foreign source text, they often use a translation tool online to understand the text as a whole. What you then get, however, is by no means a correctly coherent translation of the text, but a mixture of pre-translated segments that are selected entirely out of context by a machine. However, these “translations” usually do not make sense. So, anyone who thinks that translating is no longer an art and that translators are becoming increasingly superfluous as the use of computers increases in all areas is mistaken. Of course, the use of machines and their programs also makes the work of a translator much easier.
However, there is a significant difference between the work of a translator who uses computer programs as a support and the work of a fully automatic translation machine. A translation machine is designed to store as much information as possible, in this case words and phrases. This information, when retrieved, is then sorted according to certain rules and suggested to the user. The selection of the translation is based on statistics, but not on the interpretation ability of the system. Its understanding of language is based only on a stored lexicon and simplified syntax rules. With a machine translation, one cannot therefore trust that the result will consist of grammatically correct, complete sentences with correct content.
As a translator, one nevertheless likes to use the help of special machine translation applications. However, these are not online translation tools such as Google translate, but so-called CAT tools (computer assisted translation tools). These support the translator in so far as they store previous translations as a program and carry out a comparison with already existing databases for new translations. In this way, the translator is directly shown which segments he has already translated partially or exactly at an earlier point in time, and the translator also knows which areas of the source text to be translated will require more effort. All in all, this use of software has advantages not only for the translator himself, as it shows him a better calculation of time and effort, but also for the customer, as the delivery date of the finished translation can be better estimated and the cost of a translation can be kept low, depending on the actual effort of the translator. The combination of a good translator and a supporting machine program brings the desired result of a content and linguistically correct translation, which can also be produced quickly.
Critics would say that one would only have to continue to fill the machines with information so that at some point all linguistic expressions and all grammatical constructions that exist would be available in one large system and could be retrieved appropriately. However, this will not be possible because not every minute feature of our language can be documented. Language and communication are largely based on experience, ever-changing cultures, and feelings. Machines are not capable of capturing these inexpressible realms. Feelings and culture cannot be calculated or statistically classified. As long as humans produce source texts, it will not be possible for a machine alone to translate them for intelligent use and according to human requirements.