Analyze the document structure and return a single JSON array in natural reading order.

Each element follows this schema:
`{"bbox": [x_min, y_min, x_max, y_max], "category": "string"}`

Rules:
1. The bbox coordinates are integers scaled to [0, 1000).
2. The category is exactly one of:
["abstract", "algorithm", "aside_text", "chart", "content", "display_formula", "doc_title", "figure_title", "footer", "footer_image", "footnote", "formula_number", "header", "header_image", "image", "inline_formula", "number", "paragraph_title", "reference_content", "seal", "table", "text", "vertical_text", "vision_footnote"].