With the world getting more and more digitized every day, technology is taking over the common man. Neural networks applied in the Google translator are able to move from one language to another without “bridges”; although it looks like an “internal language”, it is more for relational knowledge.
Google released yet another proof that neural networks, when properly programmed and fed, can develop in somewhat unexpected ways: the Google Translator system, the Mountain View tool that many blogs out there use today to transcribe texts from Overseas sources for Arabic, have developed their own method for carrying out their tasks, in a more practical way than using English as a “bridge”.
Although many blogs are referring to the method as the creation of an “internal language” that is incomprehensible to human beings, this is not exactly what happens. It is actually closer to human thinking and how we deal with different languages in translation in JLT.
Zero-shot in translation in JLT
Google found out what its translator is doing by verifying that he was capable of what we call a “zero-shot” translation ( without any prior training ) in unforeseen cases of use, as the use of neural networks was introduced in the tool in September.
According to Google, the engine, called Google Neural Machine Translation, recognizes 103 languages and is capable of translating more than 140 billion words per day, with almost human precision. Most importantly, he learns from them. By translating “apple” as “apple” between English and Portuguese, and as “manzana” into Spanish, he will learn the meaning of that word in several languages and that way, he would link it without the need of a language doing the middle- field.
It works like this: if Google Translator is able to translate texts from English to Japanese and vice versa and also from English to Korean, in theory, the algorithm should be able to take a text in Japanese and translate it directly into the Korean and Korean to Japanese, without the Anglo-Saxon language as a bridge between the two.
Google found that the neural network is capable of this: the tool can take a sentence in one language and translate it into another even if it has not been trained to do so, because it has already learned using English as a base and now dismisses it. The researchers observed the network process (in a limited way, since the model itself is complex and not so friendly even for the developers) and noticed a curious behavior: In ‘the’ we have a representation of how the neural network of the translator appears.
A sentence translated from English into Korean and another from Japanese into English; if they have the same meaning, will be assigned the same color. In ‘b’, Google separated a specific sentence for analysis. In ‘c’, it was divided into the source languages of each of the terms. What the neural network is doing is assigning three different languages as a source to the same sentence, which in theory indicates that it has a deeper representation system for each of the terms.
Hence the concept that the algorithm has developed an “internal language” came into being, even though it is not entirely true. When we learn several languages, we do not create a single language in mind and use them all at the same time because we assign concepts and experiences to each of the words we use.
What the system is doing is going beyond just memorizing translations: He is learning the meaning of the words. He knows what an apple is and because of that, he will identify it in whatever language he is taught and will do the translations accordingly. The system absorbs, learns, memorizes and applies – as we do. He is (albeit in a very rudimentary way) learning to think without being taught to do so. And yes, it is scary.
Benefits of the Google Translator
- Easy Access
Google has become the most widely used search site in the world, so all the services it offers within its platform have a great chance of being used more than some others. This is the case with Google Translate, which is the first option that appears when we search for this type of tool within Google.
For this reason, anyone who uses the Google search engine and types in the word translator will have Google Translate as their first option. Therefore, it becomes very accessible for someone who needs translation support, and in addition, it is completely free.
2. Ease to be used
Its system is very simple to use; it does not present any difficulty even for those who have never used this service before. To begin with, unlike some other translators, Google Translate has a clean image, does not present a lot of mixed information, which can cause some confusion for the user. Also, its main page is already in the language of the corresponding country, which also facilitates its handling.
To be able to translate a word on your page, just follow a few simple steps. First, you need to connect to the internet and enter a search page, then type Google Translate. As soon as the main page opens, just type in the word or text you want to translate. Instantly, the translation appears without having to click on a command. Now to make it even easier, as soon as you start writing the term you need to translate, it already offers you some suggestions of what you might be wanting.
These benefits are more than essential for both a common man as well as a professional translator. The Google translator neural network has performed better than was expected. People all around the world use it every day to get their tasks done.
However, no matter how good a neural network is developed, it can never beat the human brain. It is always better to get professional translation in JLT services!