Mandarin translation quality is significantly lower than other languages in Azure Translator

Hi Microsoft team,

We are currently using Azure Translator as part of our live church translation app, which supports multiple output languages including Spanish, Portuguese, Indonesian, Arabic, and Mandarin.

While the service performs very well in some languages (particularly Spanish and Portuguese), delivering accurate, contextually appropriate translations, we have received repeated negative feedback regarding Mandarin output. Users have reported that:

• Mandarin translations often appear literal or word-for-word, rather than conveying the overall meaning of sentences;
• Context and nuance are frequently lost, resulting in phrases that are awkward, incorrect, or confusing to native speakers;
• The same input audio produces far more accurate and natural-sounding output in other languages, such as Spanish and Portuguese.

This disparity in translation quality is impacting user experience and trust in our product. As this is a paid service, we want to ensure the translation quality across all supported languages — particularly Mandarin — meets the high standards we’ve observed in others under identical conditions.

Our questions:

1. Is there a known limitation or reduced model performance for Mandarin translation in Azure Translator?
2. Are there specific configuration options, best practices, or customisation features we should be leveraging to improve Mandarin output?
3. Is there any way to report examples or provide training data to help improve results for this language?

We would appreciate any insights, escalation paths, or guidance on how to address this, as Mandarin is a key language for our user base.

Thank you in advance for your help,

Lauren
Church Translator App Team