Problem with upload translation file in Azure AI Custom Translator

George-Marius BACIU 0 Reputation points
2026-02-26T09:36:58.2066667+00:00

Hello,

I am trying to use Azure AI Custom Translator. For this purpose, I have created a project and I tried to upload a dictionary file in xlf format. The upload fails and the error message is non-comprehensible.

Error_500_AzureCustomTranslator

I tried to select as a Child tag "Azure AI Custom Translator", nowhere to be found!

Where is the appropriate child tag?

Azure AI Custom Vision
Azure AI Custom Vision

An Azure artificial intelligence service and end-to-end platform for applying computer vision to specific domains.


2 answers

Sort by: Most helpful
  1. Manas R Mohanty 17,270 Reputation points Moderator
    2026-03-27T17:10:14.8466667+00:00

    Hi George-Marius BACIU

    We got below review from product group.

    Root Cause

    The .XLF file contained <trans-unit> entries where <source> and/or <target> tags were present but had no text content. The upload service requires both tags to be non-empty and rejects files that fail this check. These empty tags are valid XML so they won't surface as a schema error, making them easy to miss without targeted inspection. Especially given the size of the file (Over 800,000 lines).

    Resolution

    Parsed the file programmatically to identify and remove all <trans-unit> entries with empty <source> or <target> tags. The cleaned file uploaded successfully.  

    The customer should ensure that the XLF files being uploaded in the future are formatted correctly and contains data in the source and target tag.

    Status: working XLF file shared to customer via DFM

    Thank you.

    Was this answer helpful?

    0 comments No comments

  2. Anonymous
    2026-02-26T10:33:53.9266667+00:00

    Hi George-Marius BACIU

    When uploading an XLIFF dictionary to Azure AI Custom Translator, the portal prompts you to select a "Child tag" which identifies the XML element containing your phrase units. This isn't labeled “Azure AI Custom Translator”—you choose based on your XLF version:

    • For XLIFF 2.0 files (uses <unit> elements): choose unit as the child tag.
    • For XLIFF 1.2 files (uses <trans-unit> elements): choose trans‑unit as the child tag.

    That ensures the upload parser recognizes and correctly interprets your phrase records

    A few things to double-check:

    1. File size : Custom Translator caps individual files at 100 MB. If yours is larger, split it into chunks or ZIP multiple sets.
    2. Encoding/invalid characters: Remove any U+FFFD replacement characters—they indicate failed encoding conversions.
    3. XLIFF formatting quirks: Ensure your <seg> and </seg> tags stay on the same line (no stray newlines). An extra line break inside a segment can break the parser.
    4. Dictionary-specific checks

    Phrase/sentence dictionaries require an equal number of source and target segments. Export each column to plain text to verify alignment before upload.

    Supported language codes :

    Both source and target codes must be supported. One side has to be en-US or en-GB.

    If you’re still stuck, try uploading a small sample XLF (e.g. 10–20 entries) to see if it succeeds—this will help isolate whether it’s file size/format or something else.

    Regarding the “child tag” on Microsoft Q&A:

    after choosing the parent category “Azure AI,” type “Custom Translator” into the child-tag search box and hit Enter. If that doesn’t stick, you can also use the generic “Translator” tag—for this forum, “Azure AI Custom Translator” lives under the Translator category, not as its own first-level option.

    References:

    https://learn.microsoft.com/en-us/azure/ai-services/translator/custom-translator/concepts/dictionaries

    https://github.com/MicrosoftDocs/azure-ai-docs/blob/main/articles/ai-services/translator/custom-translator/concepts/document-formats-naming-convention.md

    I Hope this helps. Do let me know if you have any further queries.

    Thankyou!

    Was this answer helpful?

    0 comments No comments

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.