A legal firm uses Azure Speech to transcribe court proceedings. The audio contains specialized legal terminology and multiple speakers. The firm has a large set of past transcripts. What should they do to improve accuracy?

Use SSML to add legal terms to the speech recognition model.

Use Custom Speech with the legal transcripts and audio.

Use Intent Recognition to identify legal terms.

Use Speech Translation to translate the proceedings into standard language.

What is the purpose of keyword recognition in Azure Speech?

To identify the language being spoken

To transcribe all speech in an audio file

To detect a specific word or phrase in an audio stream

A multinational company needs to provide real-time translation of a live presentation from English to French and German. The audience needs both spoken translation and a text transcript. The solution must minimize latency and avoid building a complex pipeline. What should the developer use?

Use Speech-to-Text in English, then use a translation API, then Text-to-Speech for each language.

Use Speech Translation with multiple target languages, outputting both text and speech.

Use separate Speech Translation services for each target language.

Use Custom Speech to create a model that outputs French and German directly.

A legal firm records depositions in English and needs a Spanish transcript for review. The audio files are long (several hours) and are uploaded nightly. You need to generate the Spanish text with minimal manual effort. What should you use?

Use the Speech SDK's intent recognition with a LUIS app that translates to Spanish.

Use batch transcription with the Speech service and specify Spanish as the target language for translation.

Use the Speech SDK's keyword recognition to detect English phrases and then translate them.

Use the Speech SDK's real-time translation with speech-to-text output.

A company is building a voice-based customer service bot. The bot must understand a wide range of user intents, but the team has limited labeled data. They also need to handle out-of-scope utterances gracefully. What is the best approach?

Use a prebuilt Conversational Language Understanding model with a fallback intent for out-of-scope utterances.

Use Speech Translation to translate user input into a standard set of commands.

Train a Custom Speech model to recognize all possible intents.

Use Keyword Recognition to detect a fixed set of intents.

Microsoft Certified:Azure AI Engineer Associate

Domain 5Objective 2

Process and Translate Speech AI-102 Practice Questions (Page 8)

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 36–40

36application · medium
A legal firm uses Azure Speech to transcribe court proceedings. The audio contains specialized legal terminology and multiple speakers. The firm has a large set of past transcripts. What should they do to improve accuracy?
Select an answer first
37foundation · easy
What is the purpose of keyword recognition in Azure Speech?
Select an answer first
38expert · hard
A multinational company needs to provide real-time translation of a live presentation from English to French and German. The audience needs both spoken translation and a text transcript. The solution must minimize latency and avoid building a complex pipeline. What should the developer use?
Select an answer first
39application · medium
A legal firm records depositions in English and needs a Spanish transcript for review. The audio files are long (several hours) and are uploaded nightly. You need to generate the Spanish text with minimal manual effort. What should you use?
Select an answer first
40expert · hard
A company is building a voice-based customer service bot. The bot must understand a wide range of user intents, but the team has limited labeled data. They also need to handle out-of-scope utterances gracefully. What is the best approach?
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.