We have often heard and read about studies that proved that bilingual humans are smarter than their one-language counterparts. The ability of the brain and its unity to become higher and more active when learning another language well. But after many experiments, it seems that humans are not the only ones who can benefit from another language.
Since computers are gradually able to speak, but the process of teaching them a coherent language, challenges the scientists in the field of artificial intelligence, because the understanding of the meaning of text written or written depending on the context appropriate
So a salesforce team created a smart way to improve many computer language programs by making it speak another language
It has been shown that the automatic machine translation system was able to teach other useful things. This is useful to improve the intelligence of computers, chatbots and automating customer support.
Salesforce researchers; By training a deep learning system, for translation between English and German. This experiment fed a multilayered neural network with a large number of translated documents and modified network parameters until the network learned to produce a good translation on its own. This system represents words using vectors, a common method of encoding and analyzing meaning in text.
The researchers then trained the bilingual network to do a variety of things: identifying the feelings of a particular section of a text, classifying different types of questions, and answering questions. They found that their pre-trained network had outperformed its network on a network that had not learned a second .