Those who have to translate a sentence or even a short text of meaning made into another language increasingly rely on online translators . What is it about exactly? Do they work? And when were they invented?
Let’s find out in this guide, in which we will analyze the concept of artificial intelligence underlying the mechanism of online translation.
- Online translators: when they were born
- The birth of the first online translator: Google Translate
- Online Translators: Do They Work?
- How neural translation works
- The limits of online translator
Online translators: when they were born
The online translator is a concept as simple as it is revolutionary.
It is basically a matter of inserting a word or phrase into a simple form, then selecting the relevant language in which you want to translate it, to have the translation within a few milliseconds.
In reality, under this simple gesture is the so-called ‘ Neural Machine translation ‘.
To understand what it is you have to go back to the 60s.
The first translations were actually really bad. If you had to translate a word, the problem did not exist, but the moment you moved on to a sentence of complete meaning, it was precisely the meaning that was lost.
A first ingenious solution was found twenty years later in Japan by Makoto Nagao. This researcher developed a translation system based on ‘Statistical Machine Translation’. He began memorizing millions of sentences and expressions in different languages, and the result was truly stunning.
The birth of the first online translator: Google Translate
It had to wait until 2006, the year in which the giant Google introduced the first real online translation tool, still to be perfected in fact.
The tool of course was Google Translate , also used by professionals in the sector such as Content Marketing Suite.
After 10 years there was the definitive leap in quality, thanks to the assumption of the concept of neural networks that we are talking about.
The translated text was therefore no longer divided into separate and sometimes conflicting sentences, but the text, mimicking brain functioning, combined the sentences into a unitary concept with a logical sense.
Online Translators: Do They Work?
The question at this point is whether online translators, Google aside, really work and can be used constructively by anyone.
The answer is yes, in the sense that translators are designed precisely to mimic the mental processes that lead to the formulation of a concept expressed by words.
Among the advantages of these tools are:
- the wealth of vocabularies;
- instant translation in the target language without a first passage in English;
- the continuous input of idiomatic terms and sentences and the feedback of the communities that work for constant improvement.
But how exactly does this ‘neural’ translation work?
Neural translation basically processes the likelihood that certain words will occur in a certain sequence.
It is something very scientific indeed, because the nervous system proceeds in the same way.
When you listen to someone speak, you go by keywords and expect the conclusion of a sentence to respect its premises.
This is obvious for proverbs and sayings that are already structured, but also for discursive sentences.
So more than the word itself, or rather beyond the word, the network of meanings and connections of that word with the rest of the sentence is automatically translated.
After all, the translator who uses the mechanism of the neural network, makes half the errors in positioning the terms of a sentence and 20% less grammar errors.
The most astonishing thing about these translators, however, and in this sense their similarity to the human brain is most noticeable, is the ability to learn.
It may seem like a difficult concept, but in reality it is exactly what happens.
As it translates, the software stores examples of sentences that will then be used in the same contexts.
And that’s exactly what the brain does.
The limits of online translators
The limitation of an online translator is that he can never do without man.
This is because the programmer will constantly operate on the implementation of terms and phrases in the software itself.
In addition, a programmer can set up a translator to ‘switch’ the terminology depending on whether a legal document, a scientific document or a fiction document is to be translated.
Finally, depending on the sector of reference, it will be necessary for the programmer to set the exact terms, in order not to run up against unacceptable translations.
For example, if you want to translate an advertisement, you will have to know how the different languages carry the terms ‘nightgown’ or ‘lipstick’ or even ‘coffee maker’ so as not to cause confusion about the product itself.
In conclusion then, in order to better use an online translator that exploits the mechanism of neural networks, there are some useful precautions to always keep in mind to improve the quality of the translated text.
Among these there is certainly the need to type simple and coherent sentences, in which the meaning is clear.
Moreover, it is the same characteristic that the sentences pronounced in one’s mother tongue must have if one wants to be understood.