PDF Reader Assistant

Problem

PDF analysis often gets stuck in three practical cases: readable text PDFs are easy, but scanned pages, dense tables, and figure annotations need extra handling; a single research report, paper, or contract needs TOC, summary, keywords, and table extraction; and multiple documents need side-by-side comparison instead of manual copying. PDF Reader Assistant targets these engineering-side reading tasks by turning extraction, structured analysis, and multi-PDF comparison into a reusable workflow.

How it works

• Text extraction: prefers readable text, then estimates scanned content from the first 5 pages and can switch to OCR, trying pytesseract and then easyocr.
• Structured analysis: outputs TOC, head/middle/tail summary, Top 15 keywords, Markdown tables, image annotations, and stats such as total words, Chinese characters, English words, and numeric occurrences.
• Multi-PDF comparison: quickly summarizes several PDFs, then compares common keywords and key differences.
• Batch processing: scans a directory for .pdf files and produces a summary table with filename, page count, word count, Top 5 keywords, and table count.
• Large-file strategy: for documents over 50 pages, it first extracts the first 5 pages and TOC, then reads selected page ranges to avoid loading too much context at once.

Boundaries and notes

• Encrypted PDFs require a password before reading.
• Table extraction and OCR are optional enhancements; missing dependencies are skipped and noted.
• Chinese scanned-page recognition requires simplified-Chinese OCR language support.
• Extracted data is written to a temporary file and cleaned up after use.