Issue
・The production of corporate web manuals is increasing, leading to a higher workload within the company.
・There are limited external partners capable of handling AsciiDoc and Markdown, so conversion work must be done in-house.
・The entire process from writing to proofreading, HTML/PDF generation, and publication is done manually, resulting in significant labor.
Key Points
Due to the increase in production volume and the manual-centered operations,
a state was reached where an efficient and reproducible workflow had not been established.
Proposal Details
Human Science proposes an improvement plan that combines optimization of AsciiDoc operations with AI proofreading PoC development.
1. AsciiDoc Editing and Operation Design (Related Measures)
・Support for conversion from Word manuscripts and editing assistance
・Operation design based on GitHub management
2. Translation Flow Optimization
・Design a flow that connects to the translation process while keeping the AsciiDoc format
・Organize methods for utilizing glossaries and translation memories
3. Generative AI Proof of Concept (PoC) Development for Proofreading
・Develop a prototype AI proofreading tool trained on internal rules
・Implement as a VSCode extension
・Repeatedly verify and improve the accuracy of proofreading results
・Consider an operational model where AI performs the initial check and humans make the final decision
Expected Effects
The following effects are expected if this proposal is implemented.
・Reduction of work hours from editing to publication
・Establishment of a standardized production flow that is not dependent on individuals
・Improvement in proofreading efficiency and stabilization of quality
・Building a system that can maintain quality even with an increase in projects or changes in personnel
With an eye toward future automation of manual writing,
this proposal lays the foundation for practical AI utilization starting with the proofreading process.
