A call center records customer calls for quality assurance. The audio has background noise and agents speak quickly. You need to transcribe the calls accurately for search. The transcription does not need to be real time. What should you do?

Use real-time speech recognition with the standard model.

Use speech translation to convert the calls to text in another language.

Use keyword recognition to capture important phrases.

Use batch transcription with a custom speech model trained on call-center audio.

A company has a speech-to-text application that must recognize a mix of English and Spanish in the same conversation. They have a limited budget and cannot afford to train a custom model for both languages. What is the most cost-effective approach?

Use the prebuilt multilingual model that supports both languages.

Use Speech Translation to translate the conversation to English first.

Train a Custom Speech model with a mix of English and Spanish audio.

Use two separate speech-to-text services, one for each language.

A company is developing a mobile app that reads articles aloud. The app must allow users to choose between a male and female voice, and also adjust the reading speed. The developer is using Azure Speech in Foundry Tools. What should they implement?

Use Custom Speech to create a custom voice for each gender.

Use Text-to-Speech with different voice names and SSML prosody for rate.

Use Speech Translation to convert the text to audio.

Use Speech-to-Text with a voice selection parameter.

What is the purpose of intent recognition in the context of Azure Speech?

To determine what the user wants to do based on their spoken input

To translate speech into another language

Microsoft Certified:Azure AI Engineer Associate

Domain 5Objective 2

Process and Translate Speech AI-102 Practice Questions (Page 7)

Part of the Implement natural language processing solutions domain, which accounts for 15-20% of the AI-102 exam. Microsoft does not publish an official question count, but from its 100-minute exam (~40–65 total, ~6–13 in this domain), expect 2–4 from this objective — we provide 45 practice questions to prepare you well beyond it. (estimate)

45questions here

9free pages

9concepts

15-20%of the exam

Questions 31–35

31expert · hard
A call center records customer calls for quality assurance. The audio has background noise and agents speak quickly. You need to transcribe the calls accurately for search. The transcription does not need to be real time. What should you do?
Select an answer first
32expert · hard
A company has a speech-to-text application that must recognize a mix of English and Spanish in the same conversation. They have a limited budget and cannot afford to train a custom model for both languages. What is the most cost-effective approach?
Select an answer first
33foundation · easy
Which of the following is an example of an SSML element used to insert a pause in synthesized speech?
Select an answer first
34application · medium
A company is developing a mobile app that reads articles aloud. The app must allow users to choose between a male and female voice, and also adjust the reading speed. The developer is using Azure Speech in Foundry Tools. What should they implement?
Select an answer first
35foundation · easy
What is the purpose of intent recognition in the context of Azure Speech?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to AI-102

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by Microsoft. “AI-102” is a trademark of its owner, used for identification only.