Bid Document Duplication Check Expert

Problem It Solves

In bid review and self-checking, teams often need to know whether multiple bid documents contain duplicated language, overlapping personnel, contact details, bank accounts, or similar quotation tables. Manual page-by-page comparison is slow and hard to trace. This skill normalizes docx, text-based pdf, and txt bid files into structured data, then runs a collusion and duplication screening pass with a report that records the rationale behind each finding.

How It Works

The workflow is:
- Document extraction: extracts paragraphs, tables, metadata, and embedded image paths; PDFs without a text layer are marked as a limitation.
- Risk detection: prioritizes sensitive fields such as company name, project manager, contact phone, email, bank account, and unified social credit code, then checks author, creation/modification time, similar semantic paragraphs, and table structure.
- Noise reduction and grading: when the bidding document is provided, text commonly copied from the bidding document can be flagged as low-level baseline content; findings are graded High/Medium/Low with score, reason, and type.
- Report delivery: produces Markdown and docx reports covering overview, key-information collisions, semantic similarity, document attributes, table similarity, diff analysis, and conclusions.

Scope And Limitations

It is positioned as an early screening assistant, not a professional verbatim plagiarism system or legal determination tool. It supports Chinese bid documents and is best suited for 2–10 files per run. Scanned PDFs, encrypted files, and very long documents may limit coverage, so high-stakes decisions still require human review.