Exploring Latin epigraphy with Distributional Semantic Models: A pilot study

In the last few years, Distributional Semantic Models have been successfully applied to the analysis of both modern and ancient languages. In particular, Neural Language Models proved themselves to be a reliable tool to measure semantic relationships between words or documents based on their distributional properties. However, despite these achievements, up to the time of writing distributional models have not been applied to the analysis of Latin inscriptions. In this paper, we describe a pilot study on two datasets of inscriptions from Rome and Southern/Central Italy and Sardinia included in the CLaSSES database of non-literary Latin texts (http://classes-latin-linguistics.fileli.unipi.it). Our results show that the model can identify both macro-classes and subclasses of inscriptions, thus contributing to the refinement of the classification already proposed in large epigraphic databases.