Use the Automatic Speech Recognition API

You can use the Automatic Speech Recognition (ASR) APIs to run ASR evaluations programmatically to measure the recognition accuracy of your skill. Test the sample audio utterances in your audio files against ASR models and compare expected transcriptions with the actual transcriptions. For each sample utterance, the evaluation report includes accuracy metrics and pass/fail results. Use the results to resolve accuracy issues.

To learn more about ASR, see What Is Automatic Speech Recognition?.

If you want to use the ASR evaluation tool in the Alexa developer console, see ASR Evaluation Tool.

ASR implementation steps

To run an ASR evaluation, implement the APIs in the following order.

  1. To create your audio catalog and upload your audio files, call the following APIs in order:
    1. Create a catalog
    2. Associate a catalog to a skill
    3. Catalog content upload
    4. Catalog content upload
      Audio files must be .mp3, .wav, .aiff, or .ogg format.
    5. Complete Upload
    6. Get Upload Status
  2. To create an annotation set and upload the recorded utterances, call the following APIs:
    1. Create annotation set
    2. Add annotations
  3. Run the ASR evaluation.
    For details, see Run ASR evaluation.
  4. Get the ASR evaluation status and results.
    For details, see Get ASR evaluation status and Get ASR evaluation results.