Machine translation (MT) -

I’m sorry, I don’t take orders. I barely take suggestions.

Machine translation service VS. professional translation. Who’s gonna win?

Technology advances by leaps and bounds. News about machine translation service vs. professional translation flood the web. Now, are we facing a paradigm shift in terms of translation services for flesh and blood professionals?

The development of artificial intelligence attempts to eliminate existing linguistic borders. That is why language translation and interpretation services are a daily reality necessary for the advancement of globalization.

Before getting into the subject, it is important to clarify some concepts to understand how the development of artificial intelligence can affect the translation sector.

What is AI? What is a machine translation service?

Artificial intelligence is the combination of systematic operations that allow the creation of machinery with capabilities and behaviours similar to those of human beings. In other words, artificial intelligence is the endowment of human thought to a machine.

The operation of this system consists of analyzing large amounts of data to identify patterns and trends. This is how a machine translation service can make automatic predictions with great speed and precision. So, can a machine replace a meat-and-bone translator?

“Perhaps we should associate artificial intelligence as a tool and not as a substitute.”

Machine translation services facilitate your daily life

There are many machine translations services that facilitate the daily life of many people in daily or professional matters since they solve small linguistic doubts and thus help communication with the rest of the world.

Leaving aside the emotional part

Until now, these systems have been capable of automatically translating any type of text in a mechanical and technical way. The problem? They leave aside the emotional part. For this reason, a machine translation service can act as a lifesaver in some situations that do not require formal communication.

Google’s machine translation service was phrase-based, meaning the software looked for the best match for each phrase in dictionary terms.

At the end of 2016, a new method called Google Machine Neural Translation (GMNT) was created. It was based on artificial intelligence and therefore capable of reducing errors made with the previous system used by up to 80 percent.

The system is made up of an artificial neural network. It assumes each phrase as a translation unit, thus creating an automatic phrase-by-phrase context. In addition to the comparison between the translation databases already created, this machine translation service also uses a self-learning mechanism that allows you to deduce the language rules independently. That is, the neural network develops its own language by forming equivalences between phrases and words from different languages creating a much more human conceptual-semantic representation. In this way, the system creates a much more fluid translation, similar to that of a flesh and blood translator.

Language combination that works best: English to Spanish

The language combination that works best under Google Translate is English <> Spanish. This is due to the fact that the applied neural translation is much more advanced and therefore the quality of the service is much better.

It is clear that the automation of any process represents a great challenge for companies, when it is not a direct threat for those translation and interpreting companies that do not anticipate events and specialize in post-publishing.

Machine translation services not to replace translator – Proof by example

In conclusion, an automatic machine translation service can help translate but will it replace him/her entirely? Not just yet as can be seen from the example below.

Naz Mila, an Instagrammer from Turkey with nearly 900,000 followers used Google Translator and end up writing something on her body that didn’t make sense.

Naz wanted to write in ink on her skin something that had a deep meaning. She chose a well-known phrase i.e. “Only God can judge my mistakes and truths”. But things did not go as she expected.

Using Google translator, Naz translated the phrase from Turkish to English without being aware that the result that the machine translation service was giving her was not correct. “I can judge a single God with my mistakes and errors”.

Fortunately, the person responsible for making the tattoo was able to fix the error. Thank goodness!

This shows that literal translation does not take into account the context or intent of the phrase immersed in the cultural setting in which it is applied. This way, we obtain a totally erroneous translation, requiring the supervision of a human translator to review the final translations created by a machine translation service.

A Human Translator Interprets All the Possible Meanings of a Sentence

When a human translator receives a text that he must translate, he must interpret through a previous analysis all the possible meanings of each of the sentences that make up the text in a meticulous and exhaustive way. In order to carry out this work, the translator must be trained in the source and target languages to control the semantics and grammar of the text. He/she must be able to adapt it within an appropriate cultural framework.

Normally when a translator faces a text, he must know how to solve some complex problems such as:

Grammar problems: The grammar of the languages with which we work will be our main resource to know how to properly construct each of the sentences that make up a text.

Semantic problems: Those referring to the meaning or interpretation of linguistic signs such as symbols, words or expressions.
Cultural problems: They are problems related to expressions and vocabulary typical of the country of the source and destination language, such as festivals, cultural references, etc.
Syntactic problems: It refers to those concordance and hierarchical relationships between words when they are grouped and make up simple sentences or compound sentences.
Intentional problems: The intention of the text that is translated to interpret the text, for example, before an ironic phrase.
Language problems: Whatever language we immerse ourselves in, we come across words or expressions from a different language. The mix of languages means that the translator has to know the intention of the author and keep the expression in the chosen language.

What will happen to the translation industry?

What the future dictates is that machine translation services will continue to improve. They will allow the use of this work system in language translation and interpretation companies, as this is happening now with computer-assisted translation tools.

It is not a question of rejecting this work methodology, but rather of adapting it as best as possible to the processes of each of the companies creating cohesion between the machine translation service and human translation.

Although machine translation shows signs of constant improvement, it does not mean that the end result is better than that of a human translator, especially if heterogeneous samples are taken to generate patterns and create the final translations.

It is important to know that if the automatic methodology is used, the reworking and subsequent revision by a professional are essential to unify concepts and solve those mistakes made by artificial intelligence.

Consensus needed between AI and the translation industry

From the point of view of many professionals in the sector, we agree that we must create a consensus between the world of artificial intelligence and the world of translation to turn a machine translation service into an interesting work tool, but in no case conceive it as a substitute for the translator.

We still do not know how human intelligence works, so there is still a long way to go before affirming that the neural networks of artificial intelligence processes will replace human work.

Computer-Assisted Translation & Machine Translation – Resistance is Futile

Once again, a discussion took place on technological issues – in particular, the joint use of Machine Translation and Computer-Assisted Translation or Computer-Aided Translation as translation tools used by professional translators.

Machine translation – A bit of context

The attempt to manufacture automatic translation systems (MT – Machine Translation, or AT) emerged after the Second World War. By the 1970s, it was already quite clear that no computer was able to replicate all human reasoning behind decoding, translating, and recoding texts except those texts with extremely limited structure and vocabulary.

Computer-Assisted Translation Tools Have Flourished

In the last two decades, CAT (Computer-Assisted Translation – Tools) have flourished. They are optimised databases, originally integrated with text editors such as MS Word, then stand alone translation software. Those save everything a person translates – original source text and translation – and offers tools to manage glossaries. The idea is to accelerate translation work with research and typing, reusing terms and previously translated passages.

Most Machine Translation software is free – Wanna Know Why?

As of MT programs, we all know about those, mainly the free ones (Babelfish/Altavista, Systran, and, of course, Google Translate, among others). What has always been very obvious is that they are weak, making grammatical errors and gross terminology, serving only to have a vague idea of what a site addresses, for example. Yet, no self-respecting professional would even look at such a program.

MT and CAT – A marriage of convenience

But technology has been evolving, and anyone who has been ignoring these applications for some years may have been surprised to find Google Translate to be unintended and that it is a lot smarter than we thought.

Then, one day, at a ATA Congress, a couple of translators presented a research combining MT and CAT: translation memory to store and reuse excerpts and expressions, and Google Translate for segments still to be translated, all monitored and reviewed by the translator. The results were amazing as far as speed and quality were concerned. A Presentation slide was made available, which I didn’t see at that time but I would follow some of the authors’ e-mail exchanges on the ATA mailing list. Again at the time, it was the kind of application for which I had the highest expectation.

Everybody’s Talking About It

Interestingly, almost 11 years ago, I wrote a monograph on machine translation. Part of my essay told the story and evolution of systems, and another part presented a test with software that existed at the time. This second part is totally obsolete – the system no longer exists and, frankly, the research was very weak, just a test without much scientific basis. I only left the text available on the internet until today because it is still quite cited because of the historical summary. What I find curious is that the discussion about MT is coming back in terms not very different from the ones I used in this article, after the CATs had gained their well-deserved space in history – much of which occurred well after this work.

Machine Translation Born of an Impossible Ideal

In short: MT and computer-assisted translation were born of an impossible ideal. It was relegated to a by-product that does not affect the lives of translators. Suddenly, it comes around as something that yes, can be very useful, but raises a lot of questions, professional and ethical.

If, on the one hand, there are serious research and experiments, there is also the world of ordinary people. The translation market continues to grow. There is an increasing demand and more and more people wanting to be a translator. People wanting to learn and improve themselves, and also people wanting “easy” money (note the quotes).

Everybody’s – Still – Talking About It

Because of all this, discussions about market trends are boiling over on the “new” application of free MT software. Will many unprepared people use these applications without discretion? Do they even pose a threat to the “big” translators or do they even need to worry?

You will find many discussions on this topic in the Localisation professional group on Facebook.

In my opinion, those days, we can no longer bury our heads in the sand and fail to take notice and participate in this discussion. If several tools have already led us to increase productivity, quality and consistency of translations, the pressure for high productivity will only increase with the new generation of tools. The differential translators who believe they deserve to earn well for their work will have to account for this productivity, in addition to a quality far above the market average, since productivity without much quality is easy to achieve.

Machine Translation & Computer-Assisted Translation – Resistance is Futile

The pessimists used to say that soon we would be proofreaders of the machines. Well, it is very much the case now.

Anyway, I want those machines to be my partners – it makes no sense to be afraid of them or to fight them.

“Resistance is futile”

Sauve qui peut ! Mon traducteur est en panne

Nombreuses sont les personnes qui pensent – à tort – qu’il suffit d’utiliser un logiciel de traduction automatique pour faire traduire tout et n’importe quoi.

Or, ces logiciels ne sont pas toujours fiables.

Un propriétaire de restaurant en apprit amèrement la leçon lorsqu’il voulut traduire en anglais l’enseigne de son restaurant. Or, ce jour là, le serveur était en panne.

Le jour de l’ouverture, le propriétaire du restaurant affichait fièrement son enseigne, qui disait en anglais : « erreur du serveur de traduction ».

Attention, certains traducteurs (humains) font aussi des erreurs. Voir ci-dessous. Cela démontre que vous devriez toujours faire appel à un traducteur professionnel pour s’occuper de vos projets de traduction.

Campagne publicitaire infructueuse

American Airlines, l’une des compagnies aériennes américaines les plus traditionnelles, lança une campagne publicitaire «Fly in Leather» pour mettre en avant ses avions avec sièges en cuir. Jusqu’ici tout va bien, mais la traduction espagnole causa un certain inconfort. La traduction « vuela en cueros », en effet, était une invitation inhabituelle pour les clients à voler nus. Cela ne fonctionna pas, bien sûr.

La rébelle du tatouage

Parfois, ces tatouages écrits circulent sur Internet qui attirent davantage l’attention sur leurs fautes d’orthographe que sur leur manque de beauté. Si des erreurs se produisent avec des mots dans votre propre langue, imaginez ce qui peut se passer avec une traduction? Eh bien, même la célèbre chanteuse pop Rihanna en a fait les frais. La traduction anglais-français de “fleur rebelle” fût traduite par “Rebelle Fleur”, alors que le contraire aurait été correct.

Fin des négociations

Il est rapporté que dans la lointaine année de 1830, au milieu des négociations entre la France et les États-Unis, un secrétaire mit fin aux négociations entre les pays en raison d’une traduction erronée.

Le message transmis à la Maison Blanche disait “Le gouvernement français demande” mais a été mis comme “le gouvernement français exige”. La confusion est survenue parce que le verbe français «demander» fût changé en « demand » (exiger) en anglais. Le président américain n’aima pas le ton et suspendit les pourparlers. Ce n’est qu’après le malentendu résolu que les pays reprirent leurs négociations.

Comme vous pouvez le voir, une mauvaise traduction peut vite conduire à la catastrophe. Donc, lorsque vous avez besoin d’une traduction, faites appel à un traducteur professionnel.