Collecting & vetting training datasets
This service allows you to create your own brand of machine translation from open-source software toolkits. The models will run on your own hardware, fully secure, and eliminating the dependency on any third party.
What we do:
There is no per-word cost, and you can translate millions or even billions of words if needed. This setup gives you liberty to sell access to your MT models.
We use the same toolkits that power machine translation Microsoft Translator, European Commission, and World Intellectual Property Organization. They are based on Transformer technology and Deep recurrent neural networks.
Benefits
Quality:
Your MT engines will likely be the best in your specialty area. The service guarantees that the machine translation accuracy in-domain will be on the same level or better than stock Google Translate at launch. We ensure this by procuring and carefully vetting training datasets, running evaluations by linguistic experts in-country, and optimizing the model performance towards the selected subject area.
Requirements:
