Pronunciation Assessment


9.7 / 10



Service Level


Back to All Tutorials (4)

Building learning applications with the Pronunciation Assessment API

What can you build with the Pronunciation API ?

The output of the Pronunciation API gives you the flexibility to build many exciting user learning experiences, we love seeing what amazing new products our clients can build with our tech.

To get inspiration on what can be build you can check out our official demos:

For further inspiration, here is a summary of common patterns we see our clients build into their products:

Provide a self learning experience for pronunciation

You can use the pronunciation API to build an interface which:

  • Asks the user to pronounce the given sentence
  • Optionally gives them the option to listen to an example of the sentence being pronounced properly.
  • Once they have submitted their pronunciation display a sentence score, word scores and individual phone scores. Clients often choose a color coding system like red, orange, green to represent the scores for each phoneme.
  • You can also give the user the option to listen back to his recording.
  • You could also provide the user with individual recordings of what each phoneme should sound like.

This structure allows your users to:

  1. Listen to the expected pronunciation.
  2. Attempt their own pronunciation.
  3. See where they did wrong, listen and compare their pronunciation.
  4. Try again.

Build an English Pronunciation test

A common process for this is:

  1. Ask your users to submit their pronunciation for a set of content. Usually 5 - 10 sentences.
  2. Collect their pronunciation scores for each sentence and aggregate them to build an overall score for the test.
  3. Pass or fail the user on the test based on your internal score requirements.
  4. Display a report at the end of the test and provide the user with feedback on where he went wrong.

Recommend content based off your users common pronunciation mistakes

English learners often have a set of common pronunciation mistakes that they make.
By collecting the lowest scoring phonemes over one or multiple sentences your users submit; you can gain insight on which phonemes your users are struggling with. You can use this information to recommend content; either lessons guided towards improving their pronunciation on those phonemes; or more self paced learning lessons that contain those problem phonemes.

Build an open ended pronunciation assessment tool

If you want to able to assess your users pronunciation on arbitrary content you can leverage speech recognition to determine the transcript of the audio and then call our API. The process is usually:

  1. As the user an open ended content. e.g: Tell me about your day.
  2. Run the users audio through a speech recognition solution to generate the transcript.
  3. Send the audio and the transcript to our API to get a pronunciation report.

There are many options you can use for Automatic Speech Recognition.

Some paid options include:

There are also many free on device options: