Multi-modal, Bayesian-guided analysis of archaeological data to support archeological research

Supervisors: George Giannakopoulos, Xenia Charalambidou

Description:

Artificial intelligence (AI) is being extensively used across the scientific spectrum to support research. A domain that has yet to fully tap into the potential of AI tools is that of archaeology. In this project we evaluate if and how AI-assisted annotation, combining machine/deep learning methods with Bayesian guidance can help archaeologists organize (cluster and classify) multi-modal observations to study patterns related to colonization and migration, taking into account the explainability of the discovered patterns. Elaborating, we examine the use of AI in an ongoing study by Dr. Habil. Xenia Charalambidou focusing on a homogeneous context: a refuse pit largely filled with pottery products and workshop discards, within the settlement of Iron Age Chalcis. Dated to the late 8th and 7th centuries BCE, this deposit includes thousands of ceramic fragments, which are being examined through macroscopic and science-based analyses of provenance and technology. Τhe study will explore how AI can enhance this research by accelerating the identification of patterns (shapes, styles, fabrics, fired and misfired products) during a crucial century for the development of Aegean settlements and for processes of colonization and migration in the Mediterranean.

The students will be expected to create an AI/ML tool-set, supporting incremental clustering and classification of artefacts, based on explicit domain expert annotation. They are expected to examine, use and evaluate models, such as CNNs and Bayesian optimization to maximize the efficiency of the data annotation and organization.

Qualifications required: Programming (Python or Java), Machine Learning basics