Computational Methods for Determining the Similarity between Ancient Greek Manuscripts
Eddie B. Dunn, Curry I. Guinn, Bryan A. Reinicke, George Themelis Zervos · Annals of the Master of Science in Computer Science and Information Systems at UNC Wilmington · 2012
This paper describes research applying computational document classification methods to the domain of biblical paleography. Ancient manuscripts were preserved by the laborious process of hand-copying from prior versions. The scribes often made changes in spelling, word usage and syntax. Further, scribes might delete or alter passages that reflect a theology or understanding different from the scribe’s contemporary view. Additional material not found in the prior version(s) might be inserted, perhaps combining information from other sources. In this paper, we examine over a hundred different versions of a single apocryphal gospel, the Protoevangelium of James, in order to group the documents into families of related documents in order to better understand the history of the document and how it evolved over time. This research uses the computational techniques of k-means analysis, hierarchical clustering, and correspondence analysis to find similarities and differences between documents. These results are then compared to the work of Daniels and Zervos, scholars in the field of biblical paleography who have studied this gospel.