What's changed: Added per-section figures (cert-figure-retrofit). New AI-901 Chapter 5 (Domain 2: Content Understanding = multimodal information extraction (docs/forms/images/audio/video → structured), schema/field definition, prebuilt vs custom, positioning vs Document Intelligence/Vision/Speech, build flow = analyzer → schema → analyze → results, combining with RAG/agents)
5.1Overview of Azure Content Understanding
Understand the role of Azure Content Understanding—extracting structured fields/insights from diverse content (documents, forms, images, audio, video)—and the idea of a schema (fields) that defines what to extract.
Azure Content Understanding is a service that extracts needed information in structured form from diverse content—documents, forms, images, audio, and video. For example, "biller, amount, due date" from an invoice PDF; "summary, speaker, topics" from a meeting recording; "a description of what is shown" from an image. Input is messy "unstructured data," output is "structured data" that fits fields/tables—that conversion is its role. It is the central service for AI-901’s information-extraction workload.
5.1.1Supported inputs (multimodal)
- Documents / forms: extract fields (amounts, dates, parties) from invoices, receipts, contracts.
- Images: extract a description, tags, and specified fields of what is shown.
- Audio: extract transcription, summary, speakers, key points from calls/meetings.
- Video: extract scenes, summaries, and elements from footage.
5.1.2Defining the extraction schema (fields)
The heart of Content Understanding is the schema (the definition of fields to extract). You specify "from this kind of file, which fields, in what types, to extract," and the service analyzes the content and returns each field filled in. For an invoice you might define "biller (string) / total (number) / due date (date)." Beyond ready-made general extraction (prebuilt), you can define your own fields (custom) to match your forms—a key flexibility. Internally it leverages generative AI, capturing meaning even when layouts vary.
Where it fits: previously, "field extraction from forms" was Azure AI Document Intelligence, "image recognition" was Azure AI Vision, and "transcription" was Azure AI Speech—a separate service per input. Content Understanding is new in handling multimodal information extraction in one framework across these, unifying outputs via an extraction schema.
Watch out: (1) information extraction (Content Understanding: pull out structured data) is the opposite goal of generative AI (create new content). (2) Do not confuse computer vision (general image recognition) with Content Understanding (structured extraction into specified fields). (3) Extraction quality depends on a clear schema definition. (4) Mind prebuilt (generic) vs custom (for your forms).
Common: (1) "structured extraction of fields/insights from invoices/forms/images/audio/video" = Azure Content Understanding. (2) "define what to extract" = the schema (fields). (3) "extract fields specific to your forms" = custom; generic extraction = prebuilt. (4) Goal "extract" → information extraction; "create" → generative AI.
5.1.3Section summary
- Azure Content Understanding = an information-extraction service: files (docs/forms/images/audio/video) → structured data
- The core is the schema (defining extraction fields); available as prebuilt (generic) and custom (your forms)
- Information extraction (extract) is the opposite of generative AI (create); also distinct from computer vision (general recognition)
- It leverages generative AI internally and handles multimodal inputs in one framework
Sign in to track progress — Log in.
Quick check
(just a quick review)Q1. Which Azure service extracts structured fields/insights from invoices, forms, images, audio, and video?
Q2. In Content Understanding, what is the definition of "which fields, in what types, to extract" called?
Q3. Which correctly contrasts the goals of information extraction (Content Understanding) and generative AI?
Q4. To define extraction fields tailored to your own forms, which approach do you use?
Q5. To get "transcription, summary, speakers, key points" from a meeting recording, which workload/service fits?
Keep track of your progress
The full study guide is free to read. Sign up free to practice with the question bank, track what you have read, review your mistakes, and highlight passages.