Pronunciation Assessment using Python Speech SDK

Describe the bug KeyError when using Pronunciation Assessment feature in Python Speech SDK

To Reproduce

import azure.cognitiveservices.speech as speechsdk
import json

SUB_KEY="" # replace this with the subscription key
REG="" # replace this with the region key

def recognize_from_file(wav_file, pronunciation_assessment_config):
    speech_config = speechsdk.SpeechConfig(subscription=SUB_KEY, region=REG)
    speech_config.speech_recognition_language="en-US"

    audio_config = speechsdk.audio.AudioConfig(filename=wav_file)
    speech_recognizer = speechsdk.SpeechRecognizer(speech_config=speech_config, audio_config=audio_config)

    pronunciation_assessment_config.apply_to(speech_recognizer)
    speech_recognition_result = speech_recognizer.recognize_once()
    pronunciation_assessment_result = speechsdk.PronunciationAssessmentResult(speech_recognition_result)
    pronunciation_assessment_result_json = speech_recognition_result.properties.get(speechsdk.PropertyId.SpeechServiceResponse_JsonResult)
    return pronunciation_assessment_result_json

pronunciation_assessment_config = speechsdk.PronunciationAssessmentConfig(json_string="{\"referenceText\":\"good morning.\",\"gradingSystem\":\"HundredMark\",\"granularity\":\"FullText\"}")
wav_file = "" # path to a wav file
pron_assessment = recognize_from_file(wav_file, pronunciation_assessment_config)

Expected behavior code runs without error and the variable pron_assessment has the json containing the pronunciation assessment scores for the corresponding reference text and wav file.

Version of the Cognitive Services Speech SDK 1.24.2

Platform, Operating System, and Programming Language

• OS: Linux
• Hardware - x64
• Programming language: Python

Additional context

• Stack Trace:

Traceback (most recent call last):
 File "pron_assess.py", line 24, in <module>
   pron_assessment = recognize_from_file(wav_file, pronunciation_assessment_config)
 File "pron_assess.py", line 18, in recognize_from_file
   pronunciation_assessment_result = speechsdk.PronunciationAssessmentResult(speech_recognition_result)
 File "/home/azureuser/.local/lib/python3.8/site-packages/azure/cognitiveservices/speech/speech.py", line 3111, in __init__
   self._words = [PronunciationAssessmentWordResult(w) for w in nb['Words']]
 File "/home/azureuser/.local/lib/python3.8/site-packages/azure/cognitiveservices/speech/speech.py", line 3111, in <listcomp>
   self._words = [PronunciationAssessmentWordResult(w) for w in nb['Words']]
 File "/home/azureuser/.local/lib/python3.8/site-packages/azure/cognitiveservices/speech/speech.py", line 3042, in __init__
   self._accuracy_score = _json['PronunciationAssessment'].get('AccuracyScore', 0)
KeyError: 'PronunciationAssessment'

[yulin-li]

@AdityaYadavalli1 thanks for using Azure pronunciation assessment and reporting this issue, I will check.

[dargilco]

@AdityaYadavalli1 the fix is in code review now. I will close this issue. The SDK release notes of the next version (1.25) will mention a fix for this GitHub issue.