Addressing Challenges in Textual Geolocation: Insights from Hebrew Place Descriptions

Abstract This work-in-progress explores the challenges of textual geolocation in Hebrew, a morphologically rich language with limited computational resources. Using a novel crowdsourced dataset, HeGeL (Hebrew GeoLocation), we analyze 5,695 place descriptions contributed by 1,554 participants, each relying on memory-based spatial cognition. The research reveals significant challenges in geolocating Hebrew textual descriptions due to linguistic ambiguities, such as vague spatial references, cardinal direction use, and limited gazetteer matches. This study proposes advancements for Hebrew and similar low-resource languages, providing pathways to improve spatial reasoning in Natural Language Processing (NLP) tasks, emphasizing multilingual applicability and interdisciplinary collaboration.