Implement information extraction solutions 5.2 Extract content from documents

Your RAG chunker post-processes Content Understanding Markdown from insurance forms that…

Study guide skill: Produce clean, grounded representations to use with agents and RAG by using Content Understanding

Question B7-024

Your RAG chunker post-processes Content Understanding Markdown from insurance forms that contain checkboxes, equations, and barcodes, with formula and barcode extraction turned on. Which two statements describe the Markdown output? (Choose two.)

Select 2 answers.

A Checkboxes appear as the Unicode characters ☒ (selected) and ☐ (cleared) ✓ Correct answer
B Formulas are saved as images in the figures folder and linked from the Markdown
C Barcodes appear as plain-text lines that contain only the decoded value
D Formulas are written in LaTeX, inline ones enclosed in single dollar signs ✓ Correct answer
E Checkboxes appear as GitHub task-list items such as [x] and [ ]

Answer: A, D

Content Understanding produces GitHub Flavored Markdown: selection marks are encoded as ☒ for selected and ☐ for cleared, and formulas use LaTeX, with $...$ for inline and $$...$$ for display formulas. Barcodes appear as Markdown images whose alt text is the barcode kind and whose title is the value.

Deep dive

How it works

Content Understanding converts documents into GitHub Flavored Markdown, keeping content and layout for downstream use such as RAG chunking. Elements that plain text can't express get specific encodings:

Selection marks (checkboxes): Unicode characters, ☒ for selected and ☐ for cleared.
Formulas: LaTeX. Inline formulas are enclosed in single dollar signs ($...$) to keep the text flowing, display formulas use double dollar signs ($$...$$), and multiline formulas become consecutive display formulas.
Barcodes: Markdown images, with the barcode kind as alt text and the decoded value as the title, for example ![QRCode](barcodes/1.2 "https://www.microsoft.com").

Formula and barcode extraction are on by default (enableFormula and enableBarcode both default to true). For the chunker, that means: treat ☒ and ☐ as checkbox state, don't split inside a $...$ or $$...$$ span, and parse barcode values out of image titles.

The other options, and when you'd use them

Formulas saved as images: Formulas are encoded as LaTeX in the Markdown, not as figure images.
Barcodes as plain-text values: Barcodes are Markdown images whose title carries the value.
Task-list items like [x] and [ ]: Selection marks use the Unicode characters ☒ and ☐.

Exam tip

Checkbox = ☒/☐. Formula = LaTeX with $ or $$. Barcode = image with kind as alt text and value as title.

References

Last reviewed: 10 Oct 2026

Ready to test yourself? The quiz asks this and every other AI-103 question, with instant feedback. Practice this in a quiz →