Scientific Data Rescue and Reproducibility: Empirical Evaluation of Bidirectional Table Extraction and Layout Migration Between PDF, Excel, Word, and PowerPoint
Author: Julian Mercer (ORCID: 0009-0003-1162-170X) · Date: October 28, 2026 · DOI: 10.5281/zenodo.23756804
Abstract (ERIC 5-Element Standard)
The purpose of this methodological study is to evaluate the accuracy of automated tabular data extraction and bidirectional layout migration for rescuing empirical statistics trapped in legacy Portable Document Format (PDF) publications. Using a corpus of 240 scientific, governmental, and financial tables (DS-4), the methodology measures cell-level precision, recall, and F1-scores across spatial rule-line detection and whitespace heuristic algorithms during PDF-to-XLSX, XLSX-to-PDF, PDF-to-DOCX, and PDF-to-PPTX conversions. Results reveal that automated PDF-to-Excel table reconstruction achieves a mean F1-score of 0.968 across multi-column statistical layouts, while Excel-to-PDF rendering preserves 99.6 percent of cell formatting and calculated values. Compared to manual data re-entry by graduate research assistants, automated browser-based extraction reduces transcription latency by 94.2 percent and eliminates decimal-shift typographical errors. Implications for quantitative social science, educational evaluation, and meta-analytic research include a verified replication pipeline for converting archival statistical reports into machine-readable spreadsheets. Additional materials include 6 tables, 34 references, F1 evaluation equations, and 195 embedded URI links.
Statutory ERIC Topic Areas & Descriptors
Topic Areas: Information and Technology; Assessment and Evaluation; Science, Mathematics, and Environmental Education
ERIC Thesaurus Descriptors: Data Analysis; Statistical Data; Information Retrieval; Research Methodology; Spreadsheets; Reproducibility