AI-102 Question 242
Select 3You are tasked with building a multilingual meeting translation application using Azure AI Speech service. The application should capture spoken language in real-time, translate it into a target language, and output the translated speech back in the target language. Which steps should you implement to achieve this functionality?
- A
Use the Speech-to-Text feature to convert the spoken language into text and then pass the output to the Translator Text API for translation.
- B
Use the Text-to-Speech feature to convert the translated text into spoken output in the target language.
- C
Use the Custom Neural Voice feature to generate a custom voice for the translated speech output.
- D
Enable the Speech Translation feature to directly handle real-time speech-to-speech translation.
- E
Use the Speech synthesis markup language (SSML) to define voice attributes for the translated speech output.
Show answer and explanation
Correct answers: A, B, D
Explanation
To implement a real-time speech-to-speech translation solution using Azure AI Speech service, you typically use the Speech-to-Text feature to transcribe the spoken input, pass the transcription to the Translator Text API for translation, and then use Text-to-Speech to generate the translated speech. Alternatively, the Speech Translation feature can directly handle real-time speech-to-speech translation, simplifying the process. Custom Neural Voice and SSML are optional features for more advanced customization but are not required for basic translation functionality.
- A. Correct.
Correct. The Speech-to-Text feature is necessary to transcribe the spoken input into text before translation.
- B. Correct.
Correct. The Text-to-Speech service is used to convert the translated text into speech output in the desired language.
- C. Incorrect.
Incorrect. Custom Neural Voice is not required for basic speech translation functionality; it is used for creating custom voices.
- D. Correct.
Correct. The Speech Translation feature simplifies the process of translating speech-to-speech in real-time by combining multiple services.
- E. Incorrect.
Incorrect. SSML is optional and only used when you need to define specific voice attributes for speech synthesis, which is not mandatory in this scenario.