Correct answer: B
The chapter describes the audio information extraction pipeline explicitly: "Speech-to-text transcribes the audio → the transcript is processed through NER, keyword extraction, or a generative AI model to extract structured fields." Audio cannot be directly processed by OCR or image techniques. STT produces a text transcript; NER then identifies customer name, order number; field extraction maps those to the database schema. Azure Content Understanding handles this pipeline.
- AOCR reads text from images, not audio waveforms. An audio file cannot be processed by an OCR pipeline.
The AI-901 guide explains every answer.