Full-Subject Exam Paper Generator

Problem It Solves

When teaching materials, scanned notes, and past papers are scattered across PDFs, images, and scans, manual paper authoring usually requires OCR cleanup, knowledge-point extraction, template matching, and post-generation checks for answers, duplication, difficulty mix, and formatting. This skill turns a private knowledge base and a real exam template into review-ready exam papers, with structured validation before delivery.

How It Works

• OCR extraction: images and scans can use RapidOCR or EasyOCR; text-based PDF, Word, and TXT files are read directly.
• Knowledge-point mining: extracts topics, difficulty, suitable question types, and source text for traceability.
• Template cloning: parses total score, question types, item counts, point values, and structural constraints from past papers.
• Paper generation: produces 1-3 complete papers with answers, constrained to the template structure and knowledge base, then converts Markdown to .docx.
• Quality checks: quality_checker.py scores answer completeness, knowledge coverage, difficulty distribution, template match, duplication, and formatting; failed runs can be regenerated or patched.

Boundaries

Best for subject exam creation when source materials and past-paper templates exist, especially scanned or handwritten documents needing OCR. Final quality depends on source coverage, template clarity, and extracted knowledge points; for blank slates, moving exam standards, or high-stakes grading, human review of answers and point distribution is still required.