Main Subcommittee Category: Emerging Concepts and Innovative Technologies
How would you label your submission?: Industry
Abstract:
This research introduces "Geomancer," an intuitive geospatial intelligence system that combines the capabilities of Large Language Models (LLMs) with an interactive map interface. "Geomancer" supports military intelligence gathering and analysis, employing open-source and commercial LLMs to interpret and respond to queries. The user interface uses natural language and selectable areas of interest. The study evaluated "Geomancer's" efficacy in improving operational efficiency and strategic decision-making in military applications. We built an evaluation framework for "Geomancer" on three fundamental criteria: user interface friendliness, enhancement of operational productivity, and the analytical quality of geospatial data interpretation. We gauge user interface friendliness through usability studies, focusing on the system's intuitiveness, learning curve adaptability, and overall user satisfaction. We assess operational productivity enhancement by benchmarking task completion times and error rates against conventional geospatial intelligence systems. We explore the analytical quality based on the precision, contextual relevance, and depth of insights derived from the LLM with user-generated, map-based queries. The study includes military personnel and strategists engaging with "Geomancer" in simulated operational environments. We catalog and classify their interactions with the system with attention to the intuitiveness and efficiency of the interface, as well as the effectiveness of the natural language processing capabilities in handling geospatial inquiries. Feedback was obtained through structured questionnaires, complemented by objective performance data. Challenges addressed in this study included the dependency on external APIs for geospatial data and the need to continuously refine the LLM's training dataset to ensure accuracy and relevance, particularly given the dynamic nature of military intelligence requirements and social media feeds. Integrating a natural language-driven interface with a user-friendly map, "Geomancer" aims to provide a more accessible, efficient, and accurate tool for military intelligence, thereby enhancing the speed and outcomes of geographic decision-making.
Keywords: AI;GEOSPATIAL DATA ;INTERACTIVE;NATURAL LANGUAGE PROCESSING
