Redmond – Amsterdam, May 10, 2023
Microsoft and TAUS are proud to announce the launch of our latest feature for Custom Translator – domain-specific translation models! This new feature will allow customers to easily access pre-built custom models for specific industries and use cases, providing even greater flexibility and convenience for our customers.
Starting May 2023, Microsoft customers will be able to select Custom Translator domain-specific models on Azure Marketplace. The domain-specific models are provided by TAUS, a company specializing in language data and Data-enhanced MT (DeMT™). These models are pre-trained on industry-specific terminology and language, ensuring high-quality translations right out of the gate.
At launch, six custom models are available in the health and e-commerce domains. The library will be expanded to 30 models and more domains by the end of May and will continue to expand to meet customer demand.
Our goal with this new feature is to help customers save time and improve efficiency by providing ready-made solutions for their translation needs. Whether you’re working in a highly specialized industry or simply need to translate large volumes of content quickly, Custom Translator’s domain-specific models can help streamline your workflow and improve the accuracy of your translations.
“As the demand for MT continues to grow, many of our customers are interested in using out-of-the-box domain-specific models and optimize further with their domain-specific terminology and data”, says Mohamed Elghazali, Product Manager for Microsoft Translator. “However, oftentimes, customers don’t have human curated data or the skills to build their own custom engines. Now, through partnership with TAUS, we can help our customers achieve the best possible translation quality at low-cost on Microsoft Custom Translator.”
TAUS DeMT™ increases the quality of baseline MT engines by 11% to 25% (measured in BLEU scores), as shown in this evaluation report. These improvements help customers to significantly reduce the time and cost of post-editing performed by professional human translators. However, some use cases would benefit from the domain-specific MT output without human intervention.
“TAUS is specialized in language data for AI and MT”, says Dace Dzeguze, COO at TAUS. “Over the past fifteen years our team has aggregated billions of words in hundreds of language combinations. This is the primary pool of data for the creation of the domain-specific models that are now being offered to Microsoft customers. But we can go beyond that and assist customers in further optimization with both translation memory data and AI-generated data. The opportunities for enhancement and customization in this time of Large Language Models are exciting.”
With our partnership and a flexible, customizable, robust, and reliable translator platform, customers can trust Custom Translator to deliver accurate in-domain translations quickly and efficiently.
We are excited to see how our customers will leverage this new feature to improve their translation workflows and drive greater business success. Try out Custom Translator’s domain-specific offer today and experience the power and convenience of Microsoft’s leading translation platform.
TAUS was founded in 2005 as a think tank with a mission to automate and innovate translation. Ideas transformed into actions. TAUS became the language data network offering the largest industry-shared repository of data, deep know-how in language engineering and a network of HLP contributors of data around the globe. Our mission today is to empower global enterprises and their service and technology providers with data solutions that help them to communicate in all languages, faster, better and more efficiently.
About Custom Translator
Custom Translator is a feature of the Microsoft Translator service, which enables enterprises, app developers, and language service providers to build customized neural machine translation (NMT) systems. Customized translation systems seamlessly integrate into existing applications, workflows, and websites.