Chinese Patent Application Document Reviewer

Problem

Chinese patent applications often contain mismatches across the abstract, claims, and specification, along with terminology, reference numeral, or claim dependency errors. Manual review can miss cross-section issues. This skill performs rule-based checks on .docx / .doc patent files, maps issues to specific text locations, and produces a revision-ready copy.

How it works

It extracts body text, tables, and text-box content, then splits the document into abstract, abstract figures, claims, specification, and specification figures. It then applies rule-based checks such as:
- Abstract: length, paragraphing, commercial wording, reference numerals, and completeness.
- Claims: numbering, vague or absolute terms, antecedent basis, claim references, unity, and clarity.
- Specification and full text: title, sufficient disclosure, typos, terminology consistency, and support for the claims.
For actionable edits, review_adder.py applies replace or delete under Word revision mode and adds comments. For issues needing human judgment, it only adds a comment marker. A mandatory integrity validation runs before output, and a structured Markdown report summarizes findings by section.

Boundaries

It is suited for formal and consistency checks on patent applications, not as a substitute for patent attorneys or agents assessing legal risk, claim scope, or patentability. .doc conversion requires Windows, Microsoft Word, and pywin32. Footnotes and endnotes may not be fully extracted, and nonstandard patent formatting can affect section recognition and annotation placement.