In a haste to control and replicate every human thinking process or behavior, more and more companies are trying to develop and use artificial intelligence programs. As exciting for scientists as it may be, for the moment, the new developments have commercial purposes only.

Facebook, Google, Microsoft, and others have tried to incorporate into their products algorithms that perform face recognitions, or that can identify objects in an image, or that can map conversations.

In May, Google came up with SyntaxNet, a program that can read and understand human language on deeper levels.

Facebook followed the same trend and announced the release of an algorithm of its own, called DeepText. The engine will process natural language and will try to understand the meaning by setting up a contextual analysis.

DeepText should be able to handle more than 20 languages and thousands of messages per second.

The programmers explained that they used a deep learning technique which allows the system to create meaning out of words without having much pre-processing involved.

The more traditional methods relied on vast engineering and language knowledge and could not cover all the variety of slangs and spellings that are used by various people to communicate their ideas.

DeepText creators focus on obtaining knowledge that is not language dependent.

One of the challenges in the Artificial Intelligence field will be to create an algorithm that can understand even the more subtle nuances of human language, in order to decode the exact meaning of a verbal command.

This contextual understanding is now presented by Facebook as an already conquered engineering success. DeepText will allow programmers not only to understand the meaning of the words but also to figure out the context and the intent of the communication.

One of the examples given was a person who writes a post about wanting to sell his bike. DeepText would be able to extract information such as the price, the model of the bicycle, the urgency of the sale, and so on.

The principle behind the deep learning method is to create a network similar to the systems used by human thinking when it has to process information. As the human brain processes have not been completely investigated, and there is a lot of information that we don’t know on neuronal schemes, the developers had a relatively difficult time in creating a human-like algorithm.

The human thought process is implemented with small steps into computer programs. Most of these algorithms are not an exact copy of how the neuronal networks work. Developers use a simpler model that can help them make programs more efficient and more flexible – just like a fully functional human brain, only having the immense processing capacity of a computer.

Image Source: YouTube