Implement information extraction solutions 5.2 Extract content from documents
A RAG pipeline uses a custom Content Understanding document analyzer for technical…
Study guide skill: Implement analyzers for generating structured or markdown outputs for downstream reasoning by using Content Understanding
Question B7-023
A RAG pipeline uses a custom Content Understanding document analyzer for technical reports. In the Markdown output, figures appear only as . The team needs a generated description of each figure plus the chart data and diagram structure behind it. What should you configure?
Select one answer.
Answer: B
enableFigureDescription generates a natural-language description of each figure, and enableFigureAnalysis adds deeper analysis such as chart data and diagram components. With both set, the Markdown shows the image with its description, followed by a chart, table or Mermaid diagram. Either flag alone gives only one of the two, and OCR text alone already appears as the figure's alt text.
Deep dive
How it works
Content Understanding document analyzers render figures as standard Markdown images. The alt text holds the text detected inside the figure, and the rest depends on two config flags (both default to false):
- enableFigureDescription: generates a natural-language description of each figure, diagram, image or illustration. It appears as the image title.
- enableFigureAnalysis: performs deeper analysis, such as chart data extraction and diagram component identification.
The Markdown page shows what each combination looks like:
- Neither: just the image with the detected text as alt text, the
the team sees now. - Description only: the image with a generated description.
- Analysis only: the image without a description, followed by a chart, Markdown table or Mermaid diagram.
- Both: the image with its description, followed by the chart, table or Mermaid diagram.
The team needs a description plus the chart data and diagram structure, so both flags must be on. Charts come out as Chart.js JSON or a Markdown table (depending on config.chartFormat), and diagrams as Mermaid code blocks.
The other options, and when you'd use them
- Only enableFigureDescription: You get the description but no chart data or diagram structure.
- Only enableFigureAnalysis: You get the chart or diagram, but no generated description.
- enableOcr: It extracts text from images and scanned documents, and it's already on by default. That detected text is what's already showing as alt text.
Exam tip
Description = enableFigureDescription. Data and structure = enableFigureAnalysis. Need both? Set both.
References
Last reviewed: