Workflows for Creating Disasters Databases for Psinsar Postprocessing Using Natural Language Processing

Currently, there are efforts to classify and predict future displacement events (e.g., landslides, infrastructure collapse) using machine learning from PSInSAR-derived deformation maps. This paper proposes two workflows for creating databases with information about disasters from scanned documents and online news sources. The databases can be used as ground truth to confirm that the deformation profiles from the training datasets are related to actual events or as input in future machine learning algorithms.