Extraction and Visualization of Geographical Spatio-Temporal Information from Chinese Historical Newspapers

This research introduces a comprehensive framework for the extraction and visualization of geographical spatio-temporal information from Chinese historical newspapers. The framework categorizes geographical entities into five major classes and employs advanced deep learning techniques, including LSTM, BERT, CRF, Biaffine Decoder, and Boundary Smoothing, to enhance precision in extraction. Besides, geocoding, and state-of-the-art visualization tools are integrated to facilitate the display of geographical entities on interactive maps. This paper emerges as a cornerstone for enhancing the accessibility, precision, and contextual richness of historical newspapers, thereby fostering advancements in historical research, toponym resolution, toponym linkage, and the development of comprehensive toponym dictionaries.