Clustering Running Titles to Understand the Printing of Early Modern Books

We propose a novel computational approach to automatically analyze thephysical process behind printing of early modern letterpress books viaclustering the running titles found at the top of their pages. Specifically, wedesign and compare custom neural and feature-based kernels for computingpairwise visual similarity of a scanned document's running titles and clusterthe titles in order to track any deviations from the expected pattern of abook's printing. Unlike body text which must be reset for every page, therunning titles are one of the static type elements in a skeleton forme i.e. theframe used to print each side of a sheet of paper, and were often re-usedduring a book's printing. To evaluate the effectiveness of our approach, wemanually annotate the running title clusters on about 1600 pages across 8 earlymodern books of varying size and formats. Our method can detect potentialdeviation from the expected patterns of such skeleton formes, which helpsbibliographers understand the phenomena associated with a text's transmission,such as censorship. We also validate our results against a manual bibliographicanalysis of a counterfeit early edition of Thomas Hobbes' Leviathan (1651).