Neural Machine Translation (NMT): let’s go back to the origins

We have all heard of deep learning and artificial neural networks and have likely used solutions based on this technology such as image recognition, big data analysis and digital assistants that Web giants have integrated into their services.

In the last two years, much research has been conducting on applying artificial neural networks to natural language processing. Results are shared among an open source community in which SYSTRAN actively participates and shares its know-how.

A self-learning translation machine

As opposed to the current engines on the market (statistical and rule-based), a neural engine models the entire process of machine translation through a unique artificial neural network.

An Artificial Neural Network (ANN) is composed of layers of artificial neurons, the layers are connected together with weights called the parameters of the network.
A key element of neural network is in its ability to automatically correct its parameters during the training phase (few weeks). Technically, the generated output is compared to expected reference and corrective feedback sent “backward” to adjust weights and tune the network connections.

This technology which is based on complex algorithms at the forefront of Deep Learning, enables the PNMT™ (Pure Neural™ Machine Translation) engine to learn, generate the rules of a language from a given translated text and produce a translation overachieving the current state of the art and better than a non-native speaker.

Helping companies succeed in the age of real-time