Open-source PDF parser extracts layout, tables, formulas

Developer Beatriz Almeida released papero-extract, an open-source PDF parser that rebuilds reading order, tables, formulas and bounding boxes using plain geometry on CPU, with no machine learning models. It processes dense arXiv papers at a median 39 milliseconds per page with zero failures across 54 papers, runs in browsers or Python, and exports to Markdown, JSON, Word and Excel.