JOURNAL ARTICLE
A Structure-Aware Method for Document Recovery
L = Σ ‖ yᵢ − f(xᵢ) ‖²
02
RECOVERY WORKFLOW不是 OCR 文本导出，
不是 OCR 文本导出，
而是一条自动恢复流水线。Not an OCR text dump.
An automated recovery pipeline.
从上传、结构分析到编译检查和交付，全部自动完成。专用转换节点负责论文恢复，网站负责排队、预览、账户、付款和下载。Upload, structural analysis, compile checks, and delivery run automatically. Dedicated conversion nodes recover the paper while the site handles queues, preview, accounts, payment, and downloads.
PLaTeX 恢复引擎LaTeX recovery engine
恢复阅读顺序Recover reading order
重建公式与表格Rebuild math and tables
执行编译检查Run compile checks
预览已就绪Preview ready
92 / 100main.texCompilable LaTeX source
✓references.bibBibTeX references
✓figures/Recovered figure assets
✓版面解析Layout parsing
双栏顺序、标题层级、脚注、图表区域。Columns, headings, footnotes, and figure regions.
语义恢复Semantic recovery
公式、表格、引用关系和上下文纠错。Equations, tables, citations, and contextual repair.
模板适配Template adaptation
按目标期刊组织文档类、宏包和文件结构。Align document classes, packages, and project structure.
编译检查Compile verification
验证资源引用、文件完整性和 LaTeX 编译。Verify linked assets, file integrity, and LaTeX compilation.
想看具体能恢复哪些内容？Want to inspect what the system can recover?
继续看核心能力Explore capabilities →