To translate ever more documents, faster and while reducing costs, more and more companies are turning to artificial intelligence. While they seem like something out of the future, machine translation software these days is renowned for being quite reliable and accurate. Can the machine replace the translator without impacting the quality of the translation rendered? We will explain everything to you !
You will also be interested
[EN VIDÉO] Interview: how was artificial intelligence born? Artificial intelligence aims to mimic how the human brain works, or at least its logic when it comes to making decisions. Jean-Claude Heudin, director of the research laboratory of the IIM (Internet and Multimedia Institute), explains the origin of this research.
To be able to provide the most accurate translation possible, the translation software proceeds in several steps:
Source text analysis
The “source” text is the original text to be translated. The software examines it meticulously, as a human translator would: it dissects the text in order to identify the structure of each sentence, the different groups of words … It then sends all of these elements to the translation engine.
Applying the rules
In its database, the software has a multitude of semantic rules, syntax and spelling, each of which is unique to the source language. To eliminate translation problems one by one, it uses a succession ofalgorithms.
All the indices present in the sentence, such as capital letters or punctuation, allow him to translate the text as accurately as possible.
Using the dictionary
While studying every nook and cranny of the text structure, the software also uses a dictionary to find out how to translate a word that can have multiple meanings. For the most accurate translation possible, it is recommended to opt for translation software including thematic or specialized dictionaries as they act as an additional filter.
After having carefully analyzed all of the source text and having resolved the translation problems, the software performs “the transfer”. In other words, it gives its full translation of the source text, providing a target text. Once the translation is obtained, new concerns appear. Despite all the technology in the world, artificial intelligence is not (yet) able to translate certain nuances. For example, if in French it is common to say ” he has one spider on the ceiling “, The Germans prefer to say that” he has a worm in the apricot “. To avoid nonsense, a machine translation must be proofread by a human translator.
Frédéric Ibanez, CEO, Neurotrad
The role of the translator in machine translation
The translation of pictorial or idiomatic expressions, of ambiguous feelings or sentences is not yet the strong point of artificial intelligence. The more a source text will have a style richer and more worked, the more the machine is likely to make mistakes. This is why any translation resulting from a software must be verified by a translator: this is post-editing. Translation agencies like Neurotrad can offer solutions for machine translation with post-editing. The specialist pays particular attention to the theme of the text, to the style and precisely corrects all errors made by the machine (spelling, grammar and punctuation errors, etc.). Using a linguist to proofread a machine translation is therefore essential!
Faster and less expensive, machine translation software seems to be a good basis for “roughing up” the source text. However, the translator’s work remains essential to correct inaccuracies. Thanks to his expertise, the specialist guarantees a quality, reliable and intelligible translation in the language target.
Article written in partnership with Neurotrad teams
Interested in what you just read?
