GEOSPATIAL TEXT ANALYSIS ON HISTORICAL DOCUMENTS: SARAWAK GAZETTE

This study conducted geospatial text analysis on historical documents, specifically the Sarawak Gazette, a colonial-era publication in Sarawak, Malaysia. A deep-learning model successfully extracted location named entities from structured historical documents in XML format. Geospatial visualization techniques mapped the distribution of these location names across the region on the Sarawak map by utilizing advanced NLP techniques, including Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM). The analysis sheds light on the geographical context of historical places during Sarawak's colonial period, contributing to a better understanding of the historical significance and presence of specific locations. However, limitations include potential biases within the historical document collection and challenges in interpreting historical texts. Future research can enhance named entity recognition accuracy, refine location entity extraction, and explore innovative geospatial analysis methods to gain deeper insights into historical geospatial contexts. Overall, this study establishes a foundation for further investigations into geospatial analysis of historical documents, opening avenues for understanding the historical geography of Sarawak and beyond