Abstract—Born-digital PDF electronic documents might reasonably be expected to preserve useful data units of their source originals that suffice to produce executable papers for reproducible research. Unfortunately, developers of authoring tools may adopt arbitrary PDF generation strategies, producing a plethora of internal data representations. Such common information units as text paragraphs, tables, function graphs and flow diagrams, may require numerous heuristics to handle properly each vendor specific PDF file content. We propose a generic Reverse MVC interpretation pattern that enables to cope with that arbitrariness in a systematic way. It constitutes a component of a larger framework we have been developing for making executable papers out of PDF documents without injecting in the PDF file any extra data or code.

Index Terms—information retrieval models, content mining, executable papers, user interfaces

Cite: Jacek Siciare and Bogdan Wiszniewski, "Extraction of Information from Born-Digital PDF Documents for Reproducible Research," Journal of Advanced Management Science, Vol. 4, No. 3, pp. 238-244, May 2016. doi: 10.12720/joams.4.3.238-244