Share via

Pronunciation Assessment with Language en-GB- Phoneme symbols

Anju Aggarwal 0 Reputation points
2026-02-18T17:07:02.75+00:00

I am using the pronunciation assessment API for language en-GB

Doing the assessment at phoneme level

The documentation does mention this:

AccuracyScore: Phoneme level， Syllable level (en-US only)， Word level， Full Text level

I get a response with empty phoneme symbols e.g.

{"Word":"How","Offset":800000,"Duration":4100000,"Confidence":0.0,"AccuracyScore":98.0,"ErrorType":"None","Syllables":[{"Syllable":"","Offset":800000,"Duration":4100000,"AccuracyScore":76.0}],"Phonemes":[{"Phoneme":"","Offset":800000,"Duration":700000,"AccuracyScore":91.0},{"Phoneme":"","Offset":1600000,"Duration":3300000,"AccuracyScore":72.0}]},

If I can get the accuracy score, why cannot I see the phoneme symbols? Is there a setting where I can see the symbols too.

It has been working well for en-US, but now I need to support en-GB too, Kindly advise.

Azure AI Speech
Azure AI Speech

An Azure service that integrates speech processing into apps and services.

0 comments

2 answers

Sort by: Most helpful
Most helpful Newest Oldest
  1. Anju Aggarwal 0 Reputation points
    2026-02-19T13:23:22.2133333+00:00

    ok, thanks

    I already raised the feature request, not sure how soon it can be priortized.

    Thanks for your support.

    Best,

    Anju

  2. Alex Burlachenko 19,465 Reputation points Volunteer Moderator
    2026-02-19T13:00:25.04+00:00

    hi Anju Aggarwal,

    this is expected :) sorry, for pronunciation assessment, phoneme symbols are only fully supported for en-US. Even though u can request phoneme-level scoring with en-GB and u receive AccuracyScore values at phoneme level, the actual phoneme symbols are not returned for en-GB. That is why the Phoneme field is empty while timing and score are present.

    Docs hint is subtle but important syllable level is explicitly marked as en-US only, and in practice detailed phoneme symbol output is also effectively limited to en-US. For other locales like en-GB, the service can internally score at phoneme level but does not expose the IPA/phonetic symbols in the response payload. no additional setting to enable symbol output for en-GB. If u require visible phoneme symbols, the current workaround is to run assessment using en-US for symbol extraction, or separately use the lexicon/phoneme conversion features from Speech SDK to generate phonetic transcriptions while keeping pronunciation scoring in en-GB. If phoneme symbol output for en-GB is critical for ur scenario, u should raise a feature request through Azure Support or Feedback channels, because at the moment it is a service capability limitation rather than a configuration issue.

    rgds,

    Alex

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.