IMPLEMENTATION OF ARTIFICIAL INTELLIGENCE FOR CREATING MAPS

Parinda PANNOON

Master's thesis

Abstract:

Artificial Intelligence (AI) has started revolutionizing the field of cartography and geographic information systems (GIS). AI can help in processing huge amounts of spatial data quickly and automatically, reducing manual processes in terms of analysis, and geovisualization. ChatGPT is one of the AI-powered large language models that can be utilized alternatively to create maps. However, to incorporate the AI tool with data, it's necessary to understand how to create prompts for generating the most useful results. The main objective of this study is to assess the capability of AI-generated maps from ChatGPT-4 and to compare the quality with a traditional cartographic technique. Map results are developed by Prompt Engineering, the crucial methodology of large language models that can enhance to produce high-quality outputs. The prompt details can affect the level of output quality, this study leverages two different prompt patterns which are indirect prompt as a basic pattern, Cognitive verifier and Question refinement as an advanced prompt. The results are evaluated on the number of attempts, errors, incorrect results, and map completeness to show how well ChatGPT-4 is in thematic map creation. The case study utilizes the dataset from Global Wildfire Information System (GWIS) and Fire Information for Resource Management System (FIRMS), NASA that aims to create Wildfire maps in Portugal in both static and interactive maps using Python libraries such as Geopandas and Folium. The thematic methods explored include choropleth maps, dot density maps, and graduated symbol maps. The last stage of the study is performing the map evaluation according to cartographic rules. The workflow is conducted by ArcGIS Pro, Geopandas, Folium libraries and IDE tools (e.g., Google Colaboratory, Visual Studio code). Finally, the outputs are delivered as code snippets, archive chats, and maps in HTML and PNG formats. The outputs provide a step forward in insight into how ChatGPT-4 understands thematic mapping processes and the levels of map quality that can be achieved.

Abstract:

Artificial Intelligence (AI) has started revolutionizing the field of cartography and geographic information systems (GIS). AI can help in processing huge amounts of spatial data quickly and automatically, reducing manual processes in terms of analysis, and geovisualization. ChatGPT is one of the AI-powered large language models that can be utilized alternatively to create maps. However, to incorporate the AI tool with data, it's necessary to understand how to create prompts for generating the most useful results. The main objective of this study is to assess the capability of AI-generated maps from ChatGPT-4 and to compare the quality with a traditional cartographic technique. Map results are developed by Prompt Engineering, the crucial methodology of large language models that can enhance to produce high-quality outputs. The prompt details can affect the level of output quality, this study leverages two different prompt patterns which are indirect prompt as a basic pattern, Cognitive verifier and Question refinement as an advanced prompt. The results are evaluated on the number of attempts, errors, incorrect results, and map completeness to show how well ChatGPT-4 is in thematic map creation. The case study utilizes the dataset from Global Wildfire Information System (GWIS) and Fire Information for Resource Management System (FIRMS), NASA that aims to create Wildfire maps in Portugal in both static and interactive maps using Python libraries such as Geopandas and Folium. The thematic methods explored include choropleth maps, dot density maps, and graduated symbol maps. The last stage of the study is performing the map evaluation according to cartographic rules. The workflow is conducted by ArcGIS Pro, Geopandas, Folium libraries and IDE tools (e.g., Google Colaboratory, Visual Studio code). Finally, the outputs are delivered as code snippets, archive chats, and maps in HTML and PNG formats. The outputs provide a step forward in insight into how ChatGPT-4 understands thematic mapping processes and the levels of map quality that can be achieved.

Keywords

ChatGPT-4, GeoAI, Prompt engineering, Large language models, Cartography, Thematic maps.

Jazyk práce: English

Datum vytvoření / odevzdání či podání práce: 20. 5. 2024

Obhajoba závěrečné práce

• Vedúci: RNDr. Rostislav Nétek, Ph.D.

Palacký University Olomouc

Faculty of Science

Master programme / odbor:
Geoinformatics and Cartography / Geoinformatics and Cartography