Some very interesting multidisciplinary work is coming out of the intersection of computer science and classical epigraphy. A set of techniques relating to image processing have been applied to classical Greek inscriptions in order to establish the different ‘hands’ in which Greek inscriptions were written (Panagopoulos et al 2009; Tracy and Papaodysseus 2009; see also the news article here). Given 24 high-quality images of classical inscriptions, but no other information about the artifacts whatsoever, the researchers calculated ideal forms for each letter in each inscription, and then analysed the letters from each pair of inscriptions, in order to test statistically the hypothesis that the inscriptions were made by the same writer. The results show 100% agreement with the opinion of Stephen Tracy, the epigraphist associated with the study (who selected the inscriptions but had nothing to do with the image analysis), and apparently with several other epigraphists. Four of the 24 inscriptions were in fact halves of the same inscription, and in both these cases the identification of the writer was correct.
It remains to be seen how widely this technique can be applied; the Greek classical inscriptions are highly regular and the signs are not normally ligatured to one another, while a cursive script would present significantly greater difficulties. It also doesn’t prove that these were written by six individuals – for instance, if two individuals wrote at the same place and the same time in statistically indistinguishable ways, they would be grouped together. This method has equalled expert opinion on a limited corpus, and confirmed these experts’ analysis, but it has not exceeded it. Ideally we would like to be able to apply this to texts in known hands and then to use this to identify the hand of inscriptions whose authorship is completely unknown, or controversial. If in a larger test, it took a batch of inscriptions and put inscriptions thought to be the work of one writer into two different groups, that would not be a refutation of the method – it could in fact suggest that the method is more capable than the epigraphists! While more testing is necessary, this could well prove to be a major advance, not only in Greek epigraphy but in the analysis of all sorts of ancient and modern scripts.
Panagopoulos, Michail, Constantin Papaodysseus, Panayiotis Rousopoulos, Dimitra Dafi, and Stephen Tracy. 2009. Automatic Writer Identification of Ancient Greek Inscriptions. Pattern Analysis and Machine Intelligence, IEEE Transactions on 31, no. 8: 1404-1414.
Tracy, S. V., and C. Papaodysseus. 2009. The Study of Hands on Greek Inscriptions: The Need for a Digital Approach. American Journal of Archaeology 113, no. 1: 99-102.