Document Visibilty Graph Threshold Estimation Dataset

<p>The aim of this dataset is to help with an estimation of thresholds used in geometrical algorithms for the creation of Visibility Graphs out of document content.</p> <p>For that purpose, the following thresholds are optimal values, leading to an maximal Area F1 for Table Region Detection tasks where those thresholds are the basis.</p> <p>Prediction target thresholds are:</p> <ul> <li>x_eps: alignment epsilon for vertical edges in points</li> <li>y_eps: alignment epsilon for horizontal edges in points</li> <li>page_ratio_x: maximal relative horizontal distance of two nodes where an edge can be created</li> <li>page_ratio_y: maximal relative vertical distance of two nodes where an edge can be created</li> <li>threshold_page_width: Indicating at maximal which width of a node the width should be added as an edge condition</li> <li>width_pct_eps: relative width difference of nodes as a condition for vertical edges</li> <li>font_eps: Font size difference between two nodes in points, acting again as an edge condition</li> </ul> <p>Independent variables here are:</p> <ul> <li>font_size_entropy: Shannon entropy of font sizes in a document, related to if a comparision of font sizes would be meaninful or is frequently present</li> <li>font_name_entropy: Shannon entropy of font names in a document</li> <li>bold_pct: percentage of bold texts in a document</li> <li>italic_pct: percentage of italic texts in a document</li> <li>x_var: deviation of the coordinate-based horizontal differences between nodes</li> <li>y_var: deviation of the coordinate-based vertical differences between nodes</li> <li>avg_width: average width of textual elements</li> </ul> <p>The corresponding PDF documents used will be referred in upcoming versions.</p>