AI-Powered Geo-Entity Clarification Through Context-Driven Prompt Strategies

Abstract

This research introduces a groundbreaking ap- proach to resolving geospatial ambiguities within specialized domains by employing advanced generative AI and custom prompt engineering. Focused on the interpretation of oral and transcribed testimonies from Holocaust archives, the method- ology leverages cutting-edge language models (LLMs) to dis- ambiguate entities such as concentration camps, ghettos, and urban landmarks in spatially rich narratives. By utilizing in- novative zero-shot and few-shot Chain-of-Reasoning prompts in conjunction with a Retrieval-Augmented Generation (RAG) system supported by a tailored knowledge repository, the study achieves significant improvements in both accuracy and retrieval efficiency for intricate geographic entities. The findings underline the effectiveness of combining domain-aware AI techniques with contextual augmentation, paving the way for enhanced appli- cations in historical research, geographic systems, and digital humanities analysis.