A Framework to Transform Metadata and Document-Level Tabular Spatial Information and Measurements to Marine Geology Gazetteer

In many natural science domains, scientific documents provide a very important source for fundamental measurements that have been recorded at various geo-spatial locations. Manually extracting these reported measurements and their location information from these documents is extremely tedious and expensive, necessitating an automatic approach. These documents, primarily accessible in PDF format, are designed for platform-independent viewing and printing rather than presenting content in a conveniently accessible structure, which poses a significant challenge for automatic extraction. Relevant records of measurements and location information are often provided in tables that are structured quite diversely regarding the number of (nested) rows and columns, table orientation and depend on the article template and the author’s style of presentation. In this paper, we propose a framework that, to the best of our knowledge, is the first generic approach that focuses on automatic location information and measurement extraction from tables in PDF documents. Based on various publications of arbitrary format from Marine Geology, we showcase how to segregate both spatial and quantitative measurement information, which alongside respective metadata could answer numerous pertinent questions about values of SR (Sedimentation Rate) and MAR (Mass Accumulation Rate) at specific locations, who measured these values, and referenced information (identifier) to these locations. Our proposed framework also validates the extracted information from scientific publications. Finally, all the aggregated information constitutes a novel Marine Geology Gazetteer and offers an interactive visualization.