MACHINE TRANSLATION - ARE TRANSLATORS IN DANGER?

Concept for machine translation - user pushing a "translate" button on a keyboard
Over the last few decades, the translation industry has seen numerous technological advances. While computer-assisted translation (CAT) tools and translation memories are now available to professionals to increase their productivity, they are also facing a new major challenge: machine translation.

But is the translation market really threatened and have translators become an endangered species? Before questioning automated translation, let's first have a look at its advantages, focusing on the latest neural machine translation solutions (NMT) that aim to revolutionize the translation world.

Artificial Intelligence in Translation

A few years ago, several world-renowned companies launched free online translation tools, claiming to offer a service that is more efficient than ever before. These neural machine translation solutions are based on a large parallel corpus of translated texts and “learn” on the basis of artificial intelligence tools and more particularly on trained neural networks.

Although the online translation services use technologies built on computer-based learning and neural networks, the artificial intelligence behind the new machine translation engines seems to be more powerful than those of their competitors. This is due to the fact that the efficiency of a neural network depends not only on the quantity of learning material, but rather on its quality. The NMT (Neural Machine Translation) thus ensures better results as it provides contextualized translations drawn from databases and reliable sources.

Even if these tools currently only handle a limited number of languages, they are able to provide the user with the best possible translation in terms of grammar, structures and meaning, thus producing the most natural result. They also allow millions of pieces of information to be translated in real time – thanks to supercomputers. By combining this technological breakthrough with the quality and quantity of their training material, these techniques have proven to be the most effective on the market – and have achieved unprecedented results in the BLEU (Bilingual Evaluation Understudy) test, which evaluates the quality of machine translations.

However, it is obvious that even if there has been real progress in this field and even if neural machine translation produces more natural sentences and avoids more pitfalls than other tools, some errors and misinterpretations that a translator would not have made remain. Although technical documentation demonstrates the strength of such translation tools in terms of punctuation, sentence structures and tense consistency, the results tend to be much more varied when a document contains specific technical vocabulary. The automatically generated formulations also remain “clinical” and their quality is still below the accuracy and fluidity of a human translation.

Limitations of Machine Translation

Despite the significant progress already made in machine translation, these technologies are still in their infancy and it is hard to imagine that the machine could one day completely replace the human being.

Of course, machine translation tools offer certain advantages: they are affordable or even completely free, are accessible at all times, guarantee greater speed and provide many language combinations. However, they only translate words, promote literal translation and frequently do not grasp the desired effect of using a specific word or expression. In addition, automatically translated texts lack fluidity and are sometimes somewhat incomprehensible due to the poorly phrased sentences they contain.

So even though these new technologies are an effective “decoding” tool, they are unable to read between the lines, highlight important elements and transmit the details and nuances of the original text. In some cases, long sentences may be problematic and thus lead to a lack of precision in the translation. Automatic translation platforms also do not distinguish either words with different meanings or grammatical structures with a double meaning, as they do not understand the thematic or cultural context of the source text.

It is also recommended to avoid using automatic translation for specialized texts. The content of technical documents is particularly complex and translation errors can easily lead to requests for clarification or to costly and even disastrous consequences. It is a fact that these technologies will hardly match the terminology research carried out by a human translator. A translator's creativity is indeed an added value for the client. Only a human being can meet your requirements: a translation agency always aims to provide you with translations containing consistent and homogeneous terminology, in order to avoid any subsequent revision or correction work. Whilst machine translation systems can produce results that seem correct at first glance, there is nothing like human and professional translation.

Strengths of Human and Professional Translation

In the era of globalization, translation is playing an increasingly prominent role for companies seeking to succeed in an ever more competitive international market. Though human translation presents a certain weakness in terms of costs and deadlines, it undoubtedly corrects most of the drawbacks of automatic systems whose performance does not compete with that of the human brain.

With machine translation, no matter how well the meaning of the source text is reproduced, the style “sounds like translation” and the reader can feel the machine's impersonal approach. The main advantage of human translation lies in the fact that it does not only focus on words: a professional translator must work in his or her mother tongue and therefore knows the nuances of the language (cultural or any other), allowing him or her to convey a message and an entrepreneurial spirit. Translators master the cultural aspects and linguistic nuances of their own language and can thus provide a more accurate and natural translation. They can show great creativity and are constantly searching for the expression that perfectly matches the meaning to be conveyed, to such an extent that the texts seem to have been written directly in the target language.

As regards specialised technical translations, it is only possible to guarantee optimal accuracy and to give meaning to technical terms by collaborating with people specialised in the relevant field. Knowledge of the language alone is not sufficient to translate technical texts. Our working method is based on the combination of linguistic expertise and extensive specialist knowledge.

We look for an appropriate specialised translator to handle your text, according to its particular field. Naturally, we select only translators who speak the target language of your text as their native language in addition to having a sound education and many years of experience. For a correct specialist translation, it is essential that the technical content is understood and that the translator has a sound basic knowledge of the particular field in question. Correct interpretation of the texts often requires qualified engineering-level know-how with experience in the area of application concerned. Therefore, it is obvious that this work has to be carried out by a team. At PTS, that means 100 percent checking of the technical specialist translation by an engineer who has the required know-how in the particular field and is responsible for quality.

Finally, it should be emphasised that neural machine translation requires so-called “post-editing” work (i.e. verification, adaptation and correction), since it does not yield reliable results; it is unlikely that machine translation will ever offer 100% guaranteed quality. For this reason, it cannot be ruled out that human beings and machines may be associated in the future. PTS has already applied this working method in specific cases in consultation with the customer – e. g. tender documents, specifications, etc. As machine translation services without post-editing will never comply with industry-specific quality certificates (in accordance with ISO 9001 standards), it is essential to have the automatically translated contents adapted and proofread by a professional translator or specialized engineer. Otherwise, such a strategy is simply not feasible for a certified company...

Contact us now without obligation