Digital Epigraphy at the British School at Rome

Thea Sommerschield, our Ralegh Radford Rome Awardee, speaks about her research during her time at the BSR in January – March 2020.

I was once told that, when at the British School in Rome, I should try every day to have at least two serendipitous encounters before breakfast.

I am grateful to the Professor who gave me this whimsical advice back in January, for as the months of my Award flew by, I found myself taking his advice almost to the letter. I say ‘almost’, not exclusively, because making it to hall in time for breakfast could sometimes be a challenge. More seriously in fact, by late February the steady flow of visiting scholars had dried up as worldwide events unfolded and Italy responded to the COVID-19 pandemic. By mid-March, the School had shut its newly varnished doors. Social distancing measures were put into place and we too at the BSR watched in shock as Italy descended into total lockdown.

It seemed increasingly likely that the serendipitous encounters the Professor had encouraged me to seek out were but a distant memory.

This, of course, was not true. Sadly, the toll the pandemic had on my country, Italy, was very real. Today, the lockdown measures are being eased step by step, and hope for a new beginning strengthens day by day. But during those months at the BSR, and perhaps most intensely over the course of those final weeks with my fellow award holders, I met people who inspired, supported and encouraged me, both personally and academically, through some of the most momentous events of my higher education. To these people I am forever grateful.

BSR cohort, Spring 2020, depicted invading the American Academy in Rome. Photograph by Thea Sommerschield.

Between January and mid-March, I submitted my doctoral thesis in Ancient History at the University of Oxford, and I embarked on a new chapter of my academic career with a new research project using Machine Learning models to study the epigraphic cultures of the ancient Mediterranean.

My doctoral thesis investigated how migrant and indigenous communities settled in western Sicily adopted and adapted their socio-cultural identities between the late Archaic and Classical period in response to local contexts and historical circumstances. I examined whether the written and material cultures of key sites could be used as evidence of groups and individuals expressing and negotiating these identities in order to assert social requisites and priorities, going beyond common ethnic labelling approaches to the agents and processes in this region.

Throughout my doctorate I undertook extended fieldwork in Sicily, especially to collect and study the epigraphic and material evidence I would use as case studies. On the island of Mozia I examined the tophet markers, stone votives dedicated to the god Baal Hamon alongside urns containing the cremated remains of infants and small animals. At Selinunte I studied curse tablets, lead lamellae inscribed with a curse often directed against the victim’s tongue. I also visited the necropoleis of Palermo, Monte Castellazzo, Solunto and Montelepre. The resources of the BSR, its excellent library and its support in accessing the libraries of other institutes and universities in Rome were of crucial value to my research. My thesis went on to show that distinct social groups are visible in the ways the written and material cultures of western Sicily were consciously and strategically constructed, and that certain patterns of practice worked as a medium for — and a forum of — the expression, display and negotiation of socio-cultural identities.

The tophet of Mozia. Photograph by Thea Sommerschield

In parallel with my doctorate, I co-directed a research project applying machine learning techniques to the discipline of Epigraphy — the study of ancient inscribed texts. Inscriptions are often damaged over the centuries, and missing parts of the text must be restored by specialists. This is a complex and time-consuming task, albeit a rewarding one: restoring a text means getting one step closer to understanding the historical context which produced it. We developed the first ancient text restoration model which recovers missing characters from a damaged text input using deep neural networks. We named the model Pythia, after the woman who delivered the god Apollo’s oracular responses at the Greek sanctuary of Delphi. Bringing together the disciplines of ancient history and deep learning, Pythia offers a fully automated aid to the text restoration task, providing ancient historians with multiple textual restorations, as well as the confidence level for each hypothesis.

Pythia’s architecture. Image previously published in ACL Anthology.

Pythia was trained on ancient Greek inscriptions (texts written in the Greek alphabet dating between the seventh century BCE and the fifth century CE). At the BSR, I began the expansion of this project to Latin inscriptions. The first step in this pipeline is gathering, sampling and preparing the data the model will be trained on. A key part of this process was therefore rendering the data machine readable and machine actionable by pre-processing it in Python, which began in the BSR library against an impressive green and red backdrop of Loebs. Cleaning a dataset and open-sourcing it represents a valuable contribution to the current digitisation and standardisation efforts for ancient textual corpora. In these initial stages of my research I continuously updated my knowledge of the relevant background literature in both Machine Learning and Epigraphy, and met with professors of Digital Humanities and Ancient History in Rome.

It was during this period that I refined the scope, aims and impact of my postdoctoral research proposal. I intend to explore and interpret the nature, distribution and significance of discernible patterns of practice in Greek and Latin epigraphic cultures using recent advances in the field of Machine Learning. Using computational methods to track textual connections and epigraphic parallels on an unprecedentedly large scale, this project would enable the first big data analysis of Greek and Roman epigraphic habits, thereby enriching the study of the written cultures of the ancient Mediterranean and making a transformational contribution to the study of Ancient History. Once again, the library of the BSR was fundamental for undertaking the background research concerning this ambitious project’s implementation, as were the discussions with visiting scholars and guest lecturers, fellows award holders, Roman academics and the BSR staff. I am currently applying for postdoctoral positions with this project.

To conclude, even at the height of the pandemic, the BSR offered me a conducive environment for undertaking some of the most important steps and transitions in my academic career, providing me with the resources to aid my research, and fostering an environment where at least two serendipitous encounters with inspiring people before breakfast were a welcome inevitability. These people are now my friends and colleagues, with whom I’ve shared much more than a Roman lockdown.

Thea Sommerschield (Ralegh Radford Rome Awardee, 2019–20)