AI-103: Developing AI Apps and Agents on Azure

By Microsoft · Question Mode

QuestionQ51

VerifiedImplement information extraction solutions

You have a Microsoft Foundry project that contains an agent.

The agent uses a knowledge source built from documents stored in Azure Blob Storage. The documents include digitally scanned PDFs containing tables that span multiple pages.

Your current ingestion job extracts only plain text, which causes the loss of table structure, headings, and page-number metadata. Users frequently ask questions that require retrieving specific table rows across pages.

You need to configure an ingestion job for a Retrieval Augmented Generation (RAG) pipeline that:

  • Performs optical character recognition (OCR) on the scanned PDFs
  • Preserves tables and headings as structure-aware chunks
  • Stores page-number metadata with each chunk

How should you configure the ingestion job?

Explanation

Advanced data parsing for Foundry knowledge-source ingestion performs OCR (including on scanned documents), detects and merges tables that span multiple pages while restoring column headers, preserves headings and document hierarchy, and produces semantic, structure-aware chunks that carry page-number and heading metadata — exactly the combination of OCR plus structure-aware, page-attributed chunking the scenario requires. Reingesting with advanced data parsing replaces the current plain-text extraction and directly resolves the described problem. Basic OCR with page-level chunking, or storing each whole page as a single chunk, still loses row-level table structure because a whole page (or plain OCR run) is not decomposed into structure-aware units, and basic parsing with fixed-size chunking ignores document structure entirely.

Learn more

Community Discussion

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