24003 Decoding Ancient Landscapes with AI: Leveraging Large Language Models for Archaeological Spatial Analysis

Project Title: Decoding Ancient Landscapes with AI: Leveraging Large Language Models for Archaeological Spatial Analysis 

Key Words: Archaeology, Spatial Data Analysis, Data Science, Large Language Models (LLMs), Artificial Intelligence (AI)

Research Topics: GeoAI

Mentor: Bryce Peacher, University of Central Florida

Project Description:

This project will explore how Large Language Models (LLMs), particularly Google Data Science Agent, can enhance archaeological research through spatial data analysis. The focus will be on training LLMs using archaeological datasets to improve their ability to interpret and replicate common geospatial analyses in the field. By leveraging various spatial data sources, the project will develop workflows for training LLMs to process, analyze, and derive insights into past human activities. Potential analyses include predictive modeling, viewshed, cost-surface/least-cost path, hydrological/geomorphological, spatial distribution, surface modeling, remote sensing for site detection, among others. The goal is to streamline and automate complex spatial data tasks, enabling researchers to identify patterns in historical and environmental datasets more easily. The project will offer new methodologies for integrating AI into archaeological spatial analysis, paving the way for more efficient and accurate research in the future.

Tasks and Responsibilities:

  •  Identify three representative peer-reviewed articles with reproducible datasets.
  •  Ensure datasets are structured for LLM training and testing.
  •  Train Google LLM models using these datasets and validate their performance in replicating the analytical results of the chosen articles.
  •  Develop a detailed workflow for training and fine-tuning LLM models on complex geospatial and archaeological datasets.
  •  Document the replication process and report on the comparative results between manual and LLM-assisted analysis.

Minimum qualifications:

  • -Experience in geospatial technologies, archaeology, or spatial analysis. - Basic understanding of ArcGIS and Google Earth Engine.
  • Experience working with large datasets and geospatial data (e.g., DEMs, satellite imagery, LiDAR).
  • Familiarity with common spatial analysis techniques, such as viewshed, predictive modeling, or cost surface analysis.
  •  Knowledge of Python programming, especially for data analysis and spatial data handling (e.g. Pandas, NumPy, Geopandas).
  • Familiarity with data visualization tools such as Matplotlib or Plotly.
  • Understanding of machine learning, particularly in training models or working with LLMs.
  • Strong written and verbal communication skills to effectively document findings and contribute to project reports.

Term of the Project:

  • 1 year

Deliverables:

  • Preprocessed dataset compatible with the ESRI GEOAI toolbox.
  • GeoAI analysis results, including spatial pattern detection, clustering, and predictive modeling.
  • A set of visualizations (maps, charts) showcasing key insights.
  • A comprehensive report detailing findings, interpretations, and recommendations.
  • Documentation of the workflow, methodology, and tools used for reproducibility.
  • Presentation of key findings and insights to the research team and stakeholders.
  • Peer reviewed papers, presentations, workshops/tutorials related to the project.