Treffer: Development of AI-based Archaeological Pottery Analysis System: Automatic Measurement and Seriation Using Large Language Models

This study explores the automation of archaeological chronological research using artificial intelligence Large Language Models (LLMs). In archaeological research, typological classification and seriation of pottery are fundamental and important tasks for establishing relative chronology, but they have traditionally relied on manual work, requiring significant time and effort. Therefore, this study aimed to develop an automated chronological tool to address these limitations. To achieve this objective, based on the author's previous paper “A Chronological Study of Mumun Pottery in the Hoseo Region,” we sought to implement a web-based system that integrates automatic measurement, typological classification, and seriation of pottery drawings using Large Language Models such as ChatGPT and Claude. The developed system is based on JavaScript and React framework, consisting of: an automatic measurement module that automatically recognizes and measures key metric elements from pottery drawings, a typological classification module that classifies pottery types based on measured data, and a seriation module that arranges the temporal sequence of classified pottery. In practical application, this system demonstrated efficiency by completing analyses within minutes that would take several hours manually, and showed good results in terms of user convenience and consistency of results. However, along with this efficiency, we also identified limitations and challenges in implementing AI-based systems and pottery analysis. In particular, we recognized the importance of researchers' critical role in the transparent production and interpretation of results. Nevertheless, this study demonstrates that it is possible to develop new methodologies by integrating traditional archaeological research methods with cutting-edge artificial intelligence technology, which is expected to contribute to establishing archaeological research paradigms in the AI era.