
Microsoft and Azure AI to Step Up With a Translator Upgrade
Microsoft is gearing up with an upgrade to Translator and other Azure AI services via Z-code,
By
CIO Applications Europe | Monday, October 10, 2022

Microsoft is likely to upgrade its translation services in a collaboration with AI to facilitate a comprehensive translation of global languages.
FREMONT, CA: Microsoft is gearing up with an upgrade to Translator and other Azure AI services via Z-code, a newly customised artificial intelligence model offering efficient performance and quality advantages varying from the existing large-scale language models. Xuedong Huang, the Chief Technology Officer of Azure AI and a technical member of Microsoft, commented on the update that the sole purpose of this customisation is to facilitate better communication on the planet via an enhanced quality of translation and maximised support of languages. The Z-code leverages the shared linguistic elements across multiple languages through transfer learning, where knowledge is generally applied from one task to another for an enhanced quality of machine translation and various language comprehension tasks. Similarly, it enables extending capabilities beyond common languages all across the globe to underrepresented languages with the least amount of training data.
Deploying Z-code facilitates formidable progress like exploiting both transfer and multitask learning from monolingual to multilingual data for a customised state-of-the-art language model. It encompasses a varied combination of quality, performance, and efficiency to facilitate customers with an acute translation progression. The models generally deploy a Mixture of Experts approach with increased efficiency as it utilises one single portion of the model for successful task completion, unlike other architectures that activate an entire AI model to run individual requests. Moreover, the newly emphasised model permits a huge scale of model parameters while ensuring compute constancy.
Stay ahead of the industry with exclusive feature stories on the top companies, expert insights and the latest news delivered straight to your inbox. Subscribe today.
Moreover, Microsoft is leveraging NVIDIA GPUs and Triton Inference Servers in the deployment and scaling of models before production to ensure high-performance inference. Microsoft’s utilisation of Z-code models enhances common language understanding tasks like name entity recognition, text summarisation, key phase extraction, and custom text clarification all across its Azure AI services. However, marked as the first time in history, with its public demonstration of Mixture of Experts models deployment, Microsoft employs these models in power machine translation products.
This customised Z-code translation model is likely to be available all over the world where customers utilise document translation in Translator, a Microsoft Azure Cognitive Service under Azure AI. Similarly, on account of common industry metrics, Microsoft Z-code models are constantly transforming their translation quality over current production models. Whereas, the typical multilingual transfer learning approaches exhibit AI quality gains in languages with fewer direct transaction suggestions that are generally available for training. The Z-code Mixture of Experts model shows consistent gains in the largest languages and thus shifts as a reliable translator model.
More in News
Weekly Brief
I agree We use cookies on this website to enhance your user experience. By clicking any link on this page you are giving your consent for us to set cookies. More info
Be first to read the latest tech news, Industry Leader's Insights, and CIO interviews of medium and large enterprises exclusively from CIO Applications Europe
THANK YOU FOR SUBSCRIBING


