In AI-901, this topic means understanding how Microsoft Foundry can extract useful structured information from unstructured content.
Microsoft lists this under Implement AI solutions for information extraction by using Foundry, including documents/forms, images, audio, video, and lightweight apps using Content Understanding.
What is information extraction?
Information extraction means taking unstructured content and turning it into usable data.
Examples:
| Input | Extracted information |
|---|---|
| Invoice PDF | Supplier, invoice number, date, total |
| Receipt image | Merchant, items, VAT, amount |
| Audio recording | Transcript, speakers, key points |
| Video | Spoken content, visible text, objects, scenes |
| Form | Field names and values |
Main Foundry capabilities
1. Azure Content Understanding
Use this when you need to extract information from different content types, including:
- Documents
- Images
- Audio
- Video
AI-901 expects you to recognise that Content Understanding is the Foundry capability for broader information extraction.
2. Azure Document Intelligence
Use this specifically for documents and forms.
It can extract:
- Text
- Key-value pairs
- Tables
- Document structure
- Fields from invoices, receipts, forms, and cards
Microsoft describes Document Intelligence in Foundry Tools as using OCR and document understanding to extract text, tables, structure, and key/value pairs from documents.
Exam-friendly workflow
Document / image / audio / video
↓
Foundry Tool: Content Understanding
↓
Extracted structured information
↓
Lightweight app displays or stores resultsCommon exam scenarios
| Scenario | Correct capability |
|---|---|
| Extract invoice number, supplier, and total from PDFs | Document Intelligence / Content Understanding |
| Extract text and fields from a scanned form | Document Intelligence |
| Extract information from an image | Content Understanding |
| Extract transcript and insights from audio | Content Understanding |
| Extract information from video | Content Understanding |
| Build a simple app that sends a file and receives extracted fields | Lightweight app with Content Understanding |
Key distinction
| Workload | Purpose |
|---|---|
| OCR | Reads text from an image or document |
| Text analysis | Analyses existing text |
| Document Intelligence | Extracts structured data from documents/forms |
| Content Understanding | Extracts information across documents, images, audio, and video |
| Generative AI | Creates new content |
Simple memory trick
“Extract, don’t create.”
Information extraction is about pulling useful data out of content, not generating new content.
Summary
For AI-901, remember:
Content → Foundry Tool → extracted structured data → app/workflow.