|
| 1 | +--- |
| 2 | +description: OCR a general exam PDF to accessible HTML with MathML using Mathpix API |
| 3 | +--- |
| 4 | + |
| 5 | +You are going to OCR a general exam PDF file to accessible HTML with MathML using the Mathpix API. Follow these steps precisely: |
| 6 | + |
| 7 | +## Requirements |
| 8 | +- The PDF path will be provided as an argument to this command |
| 9 | +- Use Mathpix API credentials from environment variables (MATHPIX_APP_ID and MATHPIX_API_KEY) |
| 10 | +- Generate accessible HTML with: |
| 11 | + - Actual MathML markup (NOT SVG, NOT Unicode characters) |
| 12 | + - Proper ARIA attributes (role="math", aria-label with LaTeX) |
| 13 | + - Proper semantic heading hierarchy (H1 for title, H2 for sections/problems) |
| 14 | + |
| 15 | +## Workflow |
| 16 | + |
| 17 | +### Step 1: Upload PDF to Mathpix |
| 18 | +Use the Mathpix v3/pdf API endpoint to upload the PDF: |
| 19 | +```bash |
| 20 | +curl -X POST "https://api.mathpix.com/v3/pdf" \ |
| 21 | + -H "app_id: $MATHPIX_APP_ID" \ |
| 22 | + -H "app_key: $MATHPIX_API_KEY" \ |
| 23 | + -F "file=@<PDF_PATH>" \ |
| 24 | + -F 'options_json={"conversion_formats": {"html.zip": true, "mmd": true}}' |
| 25 | +``` |
| 26 | + |
| 27 | +Extract the `pdf_id` from the response. |
| 28 | + |
| 29 | +### Step 2: Check conversion status |
| 30 | +Poll the status endpoint until conversion is complete: |
| 31 | +```bash |
| 32 | +curl -X GET "https://api.mathpix.com/v3/pdf/<PDF_ID>" \ |
| 33 | + -H "app_id: $MATHPIX_APP_ID" \ |
| 34 | + -H "app_key: $MATHPIX_API_KEY" |
| 35 | +``` |
| 36 | + |
| 37 | +Wait until `"status":"completed"`. |
| 38 | + |
| 39 | +### Step 3: Download Mathpix Markdown |
| 40 | +Download the .mmd format (NOT .html, as it uses SVG): |
| 41 | +```bash |
| 42 | +curl -X GET "https://api.mathpix.com/v3/pdf/<PDF_ID>.mmd" \ |
| 43 | + -H "app_id: $MATHPIX_APP_ID" \ |
| 44 | + -H "app_key: $MATHPIX_API_KEY" \ |
| 45 | + -o /tmp/output.mmd |
| 46 | +``` |
| 47 | + |
| 48 | +### Step 4: Convert to HTML with MathML using Pandoc |
| 49 | +```bash |
| 50 | +pandoc /tmp/output.mmd -f latex -t html --mathml --standalone -o /tmp/output_mathml.html |
| 51 | +``` |
| 52 | + |
| 53 | +### Step 5: Post-process to fix Unicode and improve accessibility |
| 54 | +Use the Python script `scripts/fix_mathml.py` to: |
| 55 | +1. Replace all Unicode mathematical characters with HTML/MathML entities (→, ∞, ℂ, ε, etc.) |
| 56 | +2. Fix heading hierarchy (first H1 stays as title, subsequent H1s become H2) |
| 57 | +3. Add ARIA attributes to all math elements (role="math", aria-label with LaTeX source) |
| 58 | +4. Add lang="en" attribute to the <html> element |
| 59 | +5. Extract title from H1 and set descriptive page title (e.g., "REAL ANALYSIS GENERAL EXAM FALL 2022 - UVA Mathematics") |
| 60 | +6. Add breadcrumb navigation at the top of the page |
| 61 | +7. Add a back button navigation below the breadcrumb |
| 62 | +8. Wrap main content in <main> landmark element |
| 63 | + |
| 64 | +**Breadcrumb navigation** should: |
| 65 | +- Have aria-label="Breadcrumb" |
| 66 | +- Use ordered list with no bullets |
| 67 | +- Include: Home / Graduate / General Exams / [Current Page Title] |
| 68 | +- Mark current page with aria-current="page" |
| 69 | +- Style separators with aria-hidden="true" |
| 70 | + |
| 71 | +**Back button** should: |
| 72 | +- Link to `/graduate/generals/` |
| 73 | +- Have aria-label="Page navigation" |
| 74 | +- Be styled as a button |
| 75 | +- Include a left arrow (←) and text "Back to General Exams" |
| 76 | + |
| 77 | +**Main landmark** should: |
| 78 | +- Wrap all content from H1 to the end |
| 79 | +- Start with `<main>` before the H1 |
| 80 | +- Close with `</main>` before `</body>` |
| 81 | + |
| 82 | +Use this entity mapping: |
| 83 | +- Arrows: → (→), ← (←), ↔ (↔), ⇒ (⇒), ⇐ (⇐), ⇔ (⇔), ↦ (↦) |
| 84 | +- Set theory: ∈ (∈), ∉ (∉), ∋ (∋), ⊂ (⊂), ⊃ (⊃), ⊆ (⊆), ⊇ (⊇), ∪ (∪), ∩ (∩), ∖ (∖), ∅ (∅) |
| 85 | +- Relations: ≤ (≤), ≥ (≥), ≠ (≠), ≈ (≈), ≡ (≡), ⟂ (⊥), ∥ (∥) |
| 86 | +- Calculus: ∂ (∂), ∫ (∫), ∮ (∮), ∑ (∑), ∏ (∏), ∇ (∇), √ (√) |
| 87 | +- Logic: ∀ (∀), ∃ (∃), ¬ (¬), ∧ (∧), ∨ (∨) |
| 88 | +- Number sets: ℝ (ℝ), ℂ (ℂ), ℕ (ℕ), ℤ (ℤ), ℚ (ℚ) |
| 89 | +- Greek letters: α (α), β (β), γ (γ), δ (δ), ε (ε), η (η), θ (θ), λ (λ), μ (μ), ν (ν), π (π), σ (σ), τ (τ), φ (φ), ω (ω), Γ (Γ), Δ (Δ), Θ (Θ), Λ (Λ), Σ (Σ), Φ (Φ), Ω (Ω) |
| 90 | +- Other: ∞ (∞), × (×), ⋅ (⋅), ± (±), ∠ (∠), ⊕ (⊕), ⊗ (⊗) |
| 91 | + |
| 92 | +### Step 6: Save to final location |
| 93 | +Save the processed HTML file next to the original PDF with the same name but .html extension. |
| 94 | + |
| 95 | +### Step 7: Add HTML link to the generals page |
| 96 | +After saving the HTML file, you MUST add a link to it in `graduate/general_exams.md`: |
| 97 | + |
| 98 | +1. Read the file `graduate/general_exams.md` |
| 99 | +2. Find the line that references the PDF you just processed |
| 100 | +3. Edit that line to add `• [HTML]({{site.url}}/path/to/file.html)` after the PDF link |
| 101 | + |
| 102 | +**Example:** |
| 103 | +```markdown |
| 104 | +# Before: |
| 105 | +- [08/2022, real]({{site.url}}/graduate/exams/analysis/2022Aug_real.pdf) |
| 106 | + |
| 107 | +# After: |
| 108 | +- [08/2022, real]({{site.url}}/graduate/exams/analysis/2022Aug_real.pdf) • [HTML]({{site.url}}/graduate/exams/analysis/2022Aug_real.html) |
| 109 | +``` |
| 110 | + |
| 111 | +**Important**: The `•` (bullet point) must be used as the separator between PDF and HTML links. |
| 112 | + |
| 113 | +## Important Notes |
| 114 | +- Do NOT use the direct .html download from Mathpix - it uses MathJax SVG rendering |
| 115 | +- Always use the .mmd → pandoc → post-processing pipeline |
| 116 | +- Verify that NO Unicode mathematical characters remain in the final output |
| 117 | +- Verify proper heading hierarchy (use grep or check a sample) |
| 118 | +- The final HTML should be fully accessible with screen readers |
| 119 | + |
| 120 | +## Success Criteria |
| 121 | +The output HTML must have: |
| 122 | +✓ Actual MathML elements (<math>, <mrow>, <mi>, <mo>, etc.) |
| 123 | +✓ NO Unicode characters - all replaced with HTML/MathML entities |
| 124 | +✓ Proper heading hierarchy (H1 for title, H2 for problems/sections) |
| 125 | +✓ ARIA attributes on all math elements (role="math", aria-label) |
| 126 | +✓ lang="en" attribute on <html> element |
| 127 | +✓ Descriptive page title derived from H1 content (e.g., "TITLE - UVA Mathematics") |
| 128 | +✓ Breadcrumb navigation with aria-label="Breadcrumb" at top |
| 129 | +✓ Back button navigation with aria-label="Page navigation" below breadcrumb |
| 130 | +✓ Main content wrapped in <main> landmark element |
| 131 | +✓ Valid, well-formed HTML5 document with proper semantic structure |
| 132 | +✓ HTML file saved next to the PDF with .html extension |
| 133 | +✓ Link added to graduate/general_exams.md with • separator |
0 commit comments