DocParser: Hierarchical Document Structure Parsing from Renderings

Translating renderings (e. g. PDFs, scans) into hierarchical document structures is extensively demanded in the daily routines of many real-world applications. However, a holistic, principled approach to inferring complete structure documents missing. As remedy, we developed “DocParser”: an end-to-end system for parsing – including all text elements, nested figures, tables, and table cell structures. Our second contribution provide dataset evaluating parsing. third propose scalable learning framework settings where domain-specific data are scarce, which address by novel weak supervision that significantly improves performance. experiments confirm effectiveness our proposed supervision: Compared baseline without supervision, it mean average precision detecting entities 39.1% F1 score classifying relations 35.8%.