Microsoft Certified: Azure AI Engineer AssociateImplement natural language processing solutionsHard

A developer is building a voice-controlled application for factory workers that operates in a loud, industrial environment. The application needs to accurately transcribe short voice commands (e.g., 'stop machine', 'start conveyor') despite significant background machinery noise. The commands use specific, domain-specific terminology. What combination of Azure AI Speech features should the developer prioritize to achieve high accuracy?

  1. ANeural Text-to-Speech and Language Understanding (LUIS)
  2. BBatch Speech-to-Text and Key Phrase Extraction
  3. CCustom Speech (Acoustic Model) and Custom Speech (Language Model)
  4. DCustom Neural Voice and Speaker Diarization
Show answer & explanation

Correct answer: C. Custom Speech (Acoustic Model) and Custom Speech (Language Model)

To achieve high accuracy in a noisy, domain-specific environment, both a custom acoustic model and a custom language model are essential. A custom acoustic model (part of Custom Speech) adapts to the specific noise and speaking styles of the factory. A custom language model (also part of Custom Speech) helps the speech-to-text engine recognize domain-specific terminology more accurately. This combination directly addresses both challenges.

Why the other options are wrong

  • A. Neural Text-to-Speech generates speech, and LUIS understands intent, not transcription accuracy for specific audio/vocabulary.
  • B. Batch Speech-to-Text is for processing large audio files, and Key Phrase Extraction is for text analysis; neither directly improves real-time transcription accuracy in noisy, domain-specific contexts.
  • D. Custom Neural Voice generates speech, and Speaker Diarization identifies speakers; neither improves transcription accuracy in noise or for domain terms.

Custom Speech (Acoustic & Language Models)

Azure AI Speech features that allow training speech-to-text models to adapt to specific acoustic conditions (noise, accents) and domain-specific vocabulary, respectively.

  • Custom acoustic models improve recognition in noisy environments.
  • Custom language models enhance recognition of domain-specific words.
  • Both are crucial for high accuracy in specialized, challenging scenarios.

Memory trick: To understand factory talk, you need a custom ear for the noise AND a custom dictionary for the jargon.

More Implement natural language processing solutions questions