Construcción de una base de datos para análisis del diseño de página de documentos digitales.

Abstract

The present work waises the creation of a Data Set for the document layout analysis (DLA) by automatic annotation. The objectives of the project were met through a five-phase methodology: collection of base files, preprocessing, data extraction, cleaning of the data set and creating the data sheet.

The collection of base files considered a significant number of scientific articles from the free digital repository arxiv.com (in .tex format) by its varied content of tables, figures, images, equations, among others. The discharged elements were adapted (preprocessing), for which each section is colored to extract the information of 12 types of regions present in the document wording.

By automatic annotation, the data extraction process was carried out, where the digital image processing techniques are highlighted (e.g., arithmetic operations between images and morphological processing). The resulting data have been validated by collaborators and in relevant cases the records have been cleaned. Finally, the data set has a data sheet with relevant information in decision making for potential users such as: motivation, composition, uses, maintenance, etc.