Analyzing Images, Editing Texts: The Manchester Project

This article discusses the methodologies and tools employed in the study of the Syriac Galen Palimpsest. While it focusses on the efforts of the ongoing Manchester Project, attention is also paid to earlier and contemporary work, particularly the most recent phase of research (which can be said to have started in 2009). In this way, the Manchester Project is properly contextualised. We describe the image analysis techniques employed by the Manchester team. The challenge is to reduce the information contained in the set of multi-spectral images and enhance it where it can usefully distinguish between undertext and overtext. One can either use unsupervised or supervised dimensional reduction techniques. An unsupervised method such as principle component analysis (PCA) provides an automatic result, whereas a supervised method such as Canonical Variates Analysis (CVA) requires one to teach the system by identifying blank areas, areas with only overtext, areas with only undertext, and areas with both. Using the resulting improvements to the visibility of the undertext, the Manchester team has been able to make significant advances in identifying where its folios fit into Galen’s Book of Simple Drugs. The use of a program called SketchEngine is outlined, which permits an engagement with parallel Greek and Syriac texts and powerful searches - this is particularly useful for those folios that come from Books 6–8, for which a parallel Syriac manuscript exists. Having completed this initial stage, it became clear that around 100 folios that did not come from Books 6-8 remained to be identified. SketchEngine again has proved to be very useful in facilitating identifications of these folios. To illustrate the different challenges posed by these two distinct scenarios, examples are provided from Books 5 and 8.