Martina Astrid is an Enrichment PhD student at the Alan Turing Institute (January-June 2019) and at the University of Oxford, Faculty of Classics. Their current research focuses on computational models of language variation and change, especially applied to ancient Greek. Prior to this, they completed a BA and MA in Classics and Ancient History from the University of Pisa and the Scuola Normale Superiore.
In addition to language variation and change, they are interested in exploring a variety of linguistic issues pertaining to ancient Greek (among others, linguistic chronology and particle usage), as well as in Imperial Greek literature (Lucian of Samosata) and theories of parody in antiquity. Their research lies at the intersection of computational linguistics and traditional philology, an area where cross-disciplinary expertise and training are in constant development.
Martina Astrid's DPhil project, supervised by Dr Philomen Probert, is a computational and corpus-based analysis of formulaic structures in archaic Greek epic. During their stay at the Alan Turing Institute, they are supervised by Dr Barbara McGillivray.
Greek epic is the product of an oral tradition of which the Homeric poems are the main remnant. During the course of this tradition, devices were developed to allow for easier composition and understanding; the most important of these is formulaic language. Formulae are recurring units which allow for limited flexibility in linguistic structure and meaning: they range from noun-epithet pairs (swift-footed Achilles) to more complex phrases. Formulaic behaviour has been compared to that of idioms and other linguistic constructions; describing the mechanisms driving formulaic variation and how they compare to other forms of linguistic change, however, is notably difficult. Martina's research explores the range of variation in the structure and meaning of Greek formulae, analysing their evolution through Homeric epic and later traditions.
During their stay at The Alan Turing Institute, they will explore how semantic flexibility, i.e. the range of meanings that a linguistic construction is allowed to take, drives the productivity and continued survival of formulae. They will apply Distributional Semantics as a data-driven method to assess semantic flexibility in a language where no native speaker input can be sought. Early Greek epic, with its high density and wide range of formulaic expressions, offers an opportunity to test existing computational models within a well-studied and stable corpus. The results gained through this approach are therefore of interest for the study of language change in general, and especially for the analysis of less well-characterised material.