Semi-Automation of Bayesian Chronology Construction Using a Graph-Theoretic Approach

In the 1990s, Bayesian modelling revolutionised archaeological chronology construction. It allowed for using relative and absolute dating evidence to improve the precision of calendar dates for archaeological events. However, the process of building such chronological models is time-consuming and labour-intensive, which means often only one chronological model is considered. Furthermore, even if multiple plausible chronological models are considered, there is currently a lack of suitable statistical metrics for analysing such sets of models. Within this research, we use mathematical graph theory to manage both stratigraphic and chronological information during Bayesian chronology construction. Dye and Buck (2015) developed a graph theoretic approach to representing archaeological relative dating evidence, proposing that we might semi-automate chronology construction by managing such archaeological dating evidence using graph theory. The research presented in this thesis builds upon the initial theoretical work by Dye and Buck (2015), showcasing novel software we have produced for the semi-automation of chronology construction. We present the result of research that sought to address three objectives. The first was to quantify the quality and potential for reuse of digital dating evidence obtained during excavation and then deposited to digital repositories. This review demonstrated a distinct lack of reusable archaeological dating evidence within a prominent digital archaeological archive in the United Kingdom. This absence of reusable data is of particular concern in archaeology due to the non-repeatable nature of excavation. The research presented in this chapter was published in a peer-reviewed journal (Moody et al., 2021), in which we provide recommendations for improving the reusability of digital dating evidence in archaeology. Following this initial research, we developed novel prototype software (using mathematical graphs) to manage dating evidence that is required for Bayesian chronological modelling. Within the same software, users can visualise and manipulate their dating evidence using point-and-click functionality. In addition, users can automatically obtain calendar age estimates for archaeological events of interest simply by loading in the required dating evidence into the software. Thus semi-automating the process of chronology construction, something which is not possible within existing software for Bayesian chronological modelling. Further, the prototype software improves the process of saving the data and information used, and produced, during the chronological building process, thus improving the potential for reusability of such data and information following future archiving. Further functionality of this software allows for the rapid semi-automated construction of multiple chronological models, which we demonstrate proof-of-concept results for case studies based on existing dating evidence from archaeological excavations carried out at various sites in Europe. Further, we explore a novel application of an existing statistical methodology that allows us to collectively interpret the results of fitting multiple chronological models, each of which we deem plausible (a priori) for a given site within our analysis. Finally, we discuss how we plan to develop our prototype software in the future, ensuring that it is functional for the archaeological community. We seek to ensure that all relevant dating evidence obtained during excavation can be managed and interpreted within the software and that it enables efficient and reliable archiving of models, methods, data and results thus facilitating improved repeatability and usability of future chronology construction methods and data.