Sharper Document Extraction with Figure Descriptions

Current release version for Sharper Document Extraction with Figure Descriptions: 0.261.001

When your admins turn on Enhanced extraction, SimpleChat uses Azure AI Content Understanding to describe figures, charts, and diagrams as it processes a file. Those descriptions become searchable text, so a question about a chart can be answered from the chart itself. Workspace document rows show a badge naming which extraction engine actually ran, and why it fell back if a different one was used.

User Side

Enhanced extraction now reads charts, diagrams, and figures inside your documents and writes searchable descriptions of them, so answers can draw on pictures instead of skipping past them.

Admin Side

Enhanced extraction is configured under Search & Extract. It now runs on Azure AI Content Understanding rather than Document Intelligence Layout, which is where the generated figure, chart, and diagram descriptions come from. Auto mode upgrades to the richer engine when a document actually contains figures, so simple text files are not pushed through a more expensive path for no benefit.

Tenants already on Enhanced or Auto are migrated on upgrade rather than being silently reset. When Content Understanding cannot run, extraction falls back to Document Intelligence and records why, and users see the engine that ran plus the fallback reason on the document row. That visibility is the point: an admin diagnosing a weak answer can tell whether the document was ever read with the richer engine, instead of guessing.

Documents processed before the change keep their original extraction until they are re-extracted, so figure-heavy documents that matter should be re-run deliberately.

Why It Matters

This matters because a large share of the meaning in reports, decks, and scanned documents lives in pictures, and until now that content was effectively invisible to search.

How to Try It

1. Open Personal Workspace and upload a document that contains charts, diagrams, or scanned figures.
2. Wait for processing to finish, then expand the document row to see the extraction badge.
3. Hover the badge to see which engine ran, and the fallback reason if a different engine was used.
4. Open the document details to read the generated figure descriptions alongside the extracted text.
5. Go to Chat, ground on that document, and ask a question that can only be answered from a figure or chart.
6. If an older document was uploaded before this change, use Change Extraction to reprocess it with the newer engine.
7. If you do not see the option, ask your admin whether Enhanced extraction is enabled for your environment.

Notes

• The Sharper Document Extraction with Figure Descriptions guide belongs to the SimpleChat 0.261.001 latest-feature set.
• The gallery for this page uses release_260_enhanced_extraction_1.png, release_260_enhanced_extraction_2.png, release_260_enhanced_extraction_3.png from the app Latest Features catalog.