Today, we’re excited to announce that we have launched the new Custom Translator portal to public preview. The new version of the Custom Translator portal offers a more intuitive experience so that you can create custom translation models more easily and efficiently.

Along with the new portal, we’ve also fine-tuned the Custom Translator V2 customization training pipeline based on feedback from early adopters. The new training pipeline is showing higher quality gains for different domain-specific business use-case and dataset sizes.

About Custom Translator

Custom Translator enables you to build translation system that reflects your business, industry, and domain-specific terminology and style. Training and publishing a custom system is easy and does not require any programming skills. The customized translation system seamlessly integrates into your existing applications, workflows, and websites and is available on Azure through the same cloud-based Microsoft Text Translator API service that powers billions of translations every day.

A well-trained custom translation system provides more accurate domain-specific translations. This is because it relies on previously translated in-domain documents to learn preferred translations. Translator uses these terms and phrases in context to produce fluent translations in the target language while respecting context-dependent grammar.

New Custom Translator Portal

The new Custom Translator portal includes significant changes to make it easier to create your custom translation systems:

• UI/UX guided experience
• Embedded deep links for more details
• Tooltips for quick info
• Actionable error messages

View the new Custom Translator portal at portal.customtranslator.azure.ai.

Fine-tuned training pipeline

The new fine-tuned data pipeline makes it easier to use Custom Translator’s V2 model training. The Custom Translator V2 platform upgrade delivers significant improvements over standard Translator and the previous V1 platform in many domains. The bar chart depicts the translation quality BLEU scores for some common domains and the impact of training dataset size on translation quality.

It is important to note that actual quality improvement is dependent upon customer data quality, training dataset size, and in-domain coverage.

Learn more

For more information on using Custom Translator and the new Custom Translator portal, refer to our fully updated documentation. Our Custom Translator Quickstart also gives step by step instructions for getting started with Custom Translator.