AI-900 Practice Question: Describe Artificial Intelligence workloads and considerations

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Identifying which employee is speaking during a meeting using their voice

  Why it's wrong here

  Voice-based speaker identification focuses on determining the unique identity of a person from their vocal characteristics, such as pitch, timbre, and spectral patterns, using a speaker model rather than recognizing linguistic content. Azure's Speech service separates speaker diarization from speech recognition, where the latter transcribes all speech to text without attributing segments to specific speakers. Therefore, this option describes speaker recognition, not the core speech-to-text capability the question asks about.

• ✓

  Converting spoken audio into written text

  Why this is correct

  Speech recognition (speech-to-text) is the task of converting an audio stream of spoken language into a textual representation, using acoustic and language models to map waveforms into phonemes, then into words and sentences. Azure Speech's real-time and batch transcription APIs process continuous speech, enabling voice commands, meeting transcription, and closed captioning. This directly matches the definition of the capability in question, making it the correct answer.

• ✗

  Recognising specific wake words to activate voice assistant devices

  Why it's wrong here

  Wake word detection, also known as keyword spotting, identifies a short, predetermined phrase—for example 'Hey Cortana' or 'Alexa'—to trigger a voice assistant, using a lightweight model optimized for low latency and always-on listening. In contrast, speech recognition recognizes arbitrary, continuous streams of speech by transcribing the entire utterance into text, not just a single trigger phrase. This option is a special-purpose detection task, distinct from the broad ability to convert spoken audio into written text, and is therefore incorrect.

• ✗

  Detecting background noise in audio to improve recording quality

  Why it's wrong here

  Detecting and removing background noise is an audio signal processing task, often accomplished by spectral analysis or noise suppression algorithms, that improves the clarity or quality of the signal but does not produce any textual output. While speech recognition may benefit from cleaner audio, its intrinsic purpose is to transcribe the words spoken, not to enhance or repair the audio stream. Thus, this option describes a preprocessing step or separate audio engineering task, not the fundamental speech-to-text function.

Go deeper

Related to this question