But can this system be used in all cases? As an option, it can be used, but at times it is not profitable or efficient from a perspective of time optimization for certain types of texts.
It is no coincidence that translation remains a task performed by humans. In many cases, it can be done with the help of translation tools, but they are just that: just help.
In most cases, if it is a repetitive text that contains short, simple sentences, this type of system can be used effectively to ensure a quality result, since it is easier for machines to interpret these sentences.
An obvious example where post-editing can be used seamlessly is the translation of user manuals or instructions, where most of the sentences are short imperatives that tell the person what to do in a very simple way.
But we can not use this system if we want to translate a marketing or advertising text where metaphors, word play or puns are the order of the day.
If we are dealing with a company’s internal memos or announcements, it may not matter much that the wording is maintained and that the style is impeccable, since the text is intended to simply report something and this goal is normally achieved with automatic translation.
If, however, we are working with a literary publication or one that is geared for advertising or even tourism, where emphasis is placed on style, it is much less appropriate to use this type of system because the post-editor would have to do perform something I might call a “quasi-translation”.
Another issue that must be considered and that is crucial in our day-to-day work is all related to the finer points of grammar. The machine cannot decide what words should be capitalized or italicized; if the original document had this type of marker, it is going to be reflected in the translation but not necessarily in the correct order and, of course, there needs to be a human touch in deciding what should be capitalized or not beyond what is reflected in the original.
Therefore, with all of this in mind, it is essential to perform tests with post-editing on random parts of the text before starting work on a project. An evaluation on the result, performed by at least the lead editor of the project if not also any translators who will be working on it, in order to analyze to what extent this system is really going to be profitable or whether it is better work only with flesh-and-blood translators.
(Versión en español: /?page_id=22843)