Google Translate is already leading the world of translation services. But researchers are preparing for an upgrade, and this would be a completely open source neural network system.
Called OpenNMT (Open Source Neural Machine Translation), the latest development comes from the researchers at Harvard. It leverages the same Torch toolkit that has been used by tech giants like Facebook, Google and IBM. Further, there is Lua language support that works with Torch.
“OpenNMT is used as provided in production by major translation providers. The system is designed to be simple to use and easy to extend while maintaining efficiency and state-of-the-art translation accuracy,” reads the official description.
Just like any other translation service, OpenNMT uses two language pairs to translate texts. It works with Torch to enable memory optimisations for high-performance GPU training. Furthermore, the open source system enables sequence generation tasks such as summarisation and image-to-text generation using extensions.
OpenNMT is a perfect solution if you are looking for an out of the box solution to enable translations, without depending on a closed source solution like Google Translate. You can install its presence on a system that has a vanilla Torch. Alternatively, you can use its Docker image to build a translation solution on the cloud.
The OpenNMT code is available in a GitHub repository. It comes with a detailed documentation and live demonstration to help fulfill your deployment.