Extensive excavations at Lepcis Magna on the coast of Libya revealed large public buildings with a variety of freestanding and relief sculpture from the Roman period. Using the catalogs from the Severan buildings, I am curating a database of reliefs in order to understand the underlying structure of the panels. Using Bayesian model selection, information theory, and generative grammars, my work identifies probable organizational principles (syntax) and infers how this information may have functioned. Analysis so far suggests a regular grammar fits our data well, one of sufficiently few states to greatly reduce the amount of information carried by figures and encourage thematic and grammatical interpretation. Likewise information theoretic analysis points towards a simple system underlying most forms of relief examined, one that nears the maximal channel capacity given the number of states in the grammar.
Ongoing work on this project includes expansion of the framework to recognize other forms of generative grammar and to codify these diverse methods into a single package that can be further used and applied to other series of reliefs. Work also continues on integrating interpretive frameworks to examine semantics and the relationships between architectural, spatial, and visual grammars. The data for Severan Lepcis Magna will be made available here and on tDAR; code will be accessible as a python script and (in part) as an R script. Work has been conducted at UNC Chapel Hill with the guidance of Prof. Gates-Foster.