AI-103: Developing AI Apps and Agents on AzureDemo

By Microsoft · Question Mode

QuestionQ42

Implement information extraction solutions

You have a Microsoft Foundry project that contains an agent.

The agent's knowledge source is a set of scanned PDF troubleshooting guides stored in Azure Blob Storage. The guide pages use two-column layouts and contain tables. You use Azure Content Understanding in Foundry Tools to process the PDFs, and you plan to ingest the processed content into an index for Retrieval Augmented Generation (RAG) while storing the extracted fields for downstream automation.

Stakeholders need to be able to verify where each extracted field value originated in the original PDF, and route low-reliability extractions for manual review.

You need the Content Understanding document analyzer output to include a per-field confidence score and source grounding to locations within the source document.

What should you do?

  • A Set enableSegment to true.
  • B Provide labeled samples.
  • C Enable estimateFieldSourceAndConfidence.
  • D Configure the analyzer to use generative extraction for all fields.
Explanation

estimateFieldSourceAndConfidence is an opt-in Content Understanding analyzer property (settable at the analyzer level or per field) that, for extractive fields, adds a confidence score between 0 and 1 plus source grounding — the page number and bounding region — for each extracted field value. This is exactly what enables verifying where a value came from in the source PDF and routing low-confidence extractions to human review. enableSegment controls document segmentation rather than confidence/grounding, providing labeled samples improves extraction accuracy over time but does not itself produce confidence-and-grounding output, and configuring generative extraction for all fields changes the extraction method but does not, by itself, enable per-field confidence and source grounding.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!