Detecting and Modeling Emotional Landscapes Through Deep Learning, GPT ‐4.1, and 3D Spatial Interpolation

ABSTRACT This research provides new directions in conceptualizing emotional geographies at the landscape level using a combination of textual deep learning (DL) and GPT‐based spatial emotion detection and interpolation using 3D empirical Bayesian kriging. The resulting multidimensional raster predicts the prevalence of emotions across the landscape, with each dimension representing a different emotion. The results show the potential for both DL and GPT methods in spatial emotion detection using contextual windows, achieving high accuracy rates in most measures, though indicating the need for improvement in precision (DL) and recall (GPT). The methodology provides further direction forward in developing a geographic information system (GIS) of place that is spatially accurate and sensitive to human experience. It also contributes to the understanding of Holocaust rescue spatially and contextually.