[Authors and titles are listed at the end of the review.]
Reviewing a book on Digital Classics is a peculiar endeavour, as the research results in this area are not in the book itself, but in cyberspace, in the form of Github repositories, searchable databases, etc.[1] Evolving Perspectives in Digital Classics is aware of this, and adopts a methodological perspective that allows the reader to grasp the processes and concepts involved in the production of a variety of digital tools that investigate Graeco-Roman Antiquity. Classicists will not learn much about the ancient texts or artefacts involved; they will, however, discover many digital projects and the perspectives they promote. The book aims to present to the public a collection of cutting-edge digital projects, to reflect on these projects, and to introduce various subfields of digital humanities, such as word vectors or Geographical Information Systems.
The book is divided into five parts, each containing two chapters. This structure is somewhat artificial: it is unclear why a paper on semantics (chapter 7) is paired with another describing attempts to decipher ancient languages for which no semantic content is yet available; part III, on AI-driven literary analysis and stylometry, would be a better fit. Similarly, part V contains an article on the educational uses of digital technologies and a very theoretical paper arguing that digital methods are shifting the way we conceive of ‘‘art objects”. Despite this, the book remains a cohesive whole, due to the way the themes announced in the introduction (Linked Open Data, interoperability, user-friendliness) are interwoven in contributions, which sometimes mention the same resources from different angles.
Digital Classics tools for Ancient Geography
The first paper in the collection describes the attempt by the Digital Periegesis project to aggregate the many different types of data (geographical, mythological, historical and artistic) found in Pausanias’ Periegesis Hellados. The Digital Periegesis project argues that it is not necessary to create new “incomplete, under-utilized databases” but rather to adapt “existing resources, in particular by exploiting the cross-disciplinary scope, openness, and flexibility offered by Wikidata” (p. 19). Wikidata’s format, which abides by the principles of Linked Open Data, enabled the connection of art history databases (Arachne), geographical elements (found in online gazetteers such as Pleiades and ToposText), mythological information (from mythoskop or MANTO) and prosopographical data (e. g. Amici Populi Romani). This chapter thus forms an excellent introduction, introducing one of the main themes of the book and many of the projects that will reappear in it. An important theme is the significance of participatory research for Digital Classics. Here as later, with the “harnessing” of Wiktionary data in chapter 5, it is clear that a significant amount of data necessary to these projects was compiled by volunteers everywhere in the world.
Chapter 2 picks up the prosopographical thread from the first. The SlaVEgents database uses the SNAP ontology in order to aggregate various sources relating to ancient slaves and support a view of slaves rooted in their agency and community rather than solely their dealings with the masters. By developing an accessible and open-access platform, SlaVEgents hopes to support future work on ancient slavery, such as the writing of slave biographies.
Chapters 3 and 4 pick up the spatial thread from chapter 1. Chapter 3 details how the MANTO project uses extant databases, mostly gazetteers of ancient places, and connects them using Linked Open Data: every place is “imbued with meaning” by being connected to characters, stories, eponyms, etc. through RDF triples. This enables ancient mythological places to be conceptualised not as “dots on a map” but as “nodes” connecting stories, mythological characters, names and etymologies. The benefits of such a method are twofold: first, they enable places that are imaginary, or both real and imaginary, such as Hades or the Isles of the Blessed, to be included along with more traditional places, such as Athens or Delphi. Conversely, it helps the user realise that “real” places were always, in ancient storytelling, made mythical through their ties to mythical narrations. The architecture of LOD thus brings us closer to how space was conceptualised in Antiquity.
Chapter 4 returns to a more down-to-earth concept of space, as it uses Geographical Information Systems to visualise the notoriously complex account of spatial directions Xenophon provides in his Anabasis. This book is mined for spatial indicators, then turned into maps, which are themselves augmented with “landscape-oriented methods” such as viewshed analysis or least-cost-path analysis. These maps are interpretations of the text, and they enable us to better “track” the Ten Thousands. The conclusion of this paper brings us back to mythic space: when GIS fails to map Xenophon’s route in Cartesian space, as is the case with his account of the “Chalybians”, this failure is heuristic: it shows that Xenophon is using this people not as a geographical entity, but as a mythical unit marking the furthest extant of his travel.
Digital Philology
The next group of chapters (5 to 8) focuses on the most canonical objects of Classical Studies, the texts. Chapter 5 makes the case for a symbiotic relationship between AI (i. e. Bidirectional Encoder Representations from Transformers [BERT]) and Classics, arguing that AI can help us read ancient texts better while Classicists’ engagements with AI can make it more human-centred. The latter claim is weakly defended: manual consideration of the data and open-access, open-source documentation seems to be a low bar to pass for an AI to be called “ethical” (p. 110); however, the paper explains with clarity and concision the steps that were followed to develop a sentiment analysis BERT for ancient Greek and Latin texts: curation and uniformisation of XML files, lemmatisation, generation of random Latin sentences, unsupervised training based on sentiment dictionaries. A sample of convincing results is provided.
Chapter 6 is a very well-written introduction to one of the most ancient computational techniques applied to literature, stylometry. It sums up the basics of this approach, explains the most technical terms and techniques of supervised training (decision trees, random forest classifiers, etc.) and then provides four case studies, ranging from the most basic application (telling verse from prose) to more subtle literary questions, such as sentence-length in Latin drama, or the ways we can select the most salient aspects of tragedy, comedy or different styles of epistolography using dimensionality reduction.
Chapter 7 provides an equally clear and thorough introduction to the various types of word-vectors (static word vectors, static subword-vectors, and contextual vectors) and the unsupervised training methods they sustain. A case study shows that BERTs trained with contextual word-vectors are able to bring a subtlety to the detection of intertextual relationships in Latin hexameters that was previously available only to the patient and knowledgeable reader, as they take into account not only word and n-gram reuse, but also semantic proximity, thus bringing out similarities between lines that barely have a word in common.
Chapter 8 focuses on digital contributions to the study of rare, undeciphered scripts. It makes a less easy read than the previous chapters and assumes more knowledge of photogrammetry than most reader will probably have; however, it does give an idea of the difficulties involved in “translating” into digital images the crucial, yet minute details that make up undeciphered inscriptions (think of how relevant the depth of an incision may be on deciding whether it is a linguistic sign or a fluke line). Progresses in this domain can be seen on the INSCRIBE 3D Interactive Web viewer.
Pedagogical and epistemological conclusions
Part V of the book collects two conclusions: the first, from the perspective of Latin and Ancient Greek teachers, is a paper on the educational uses of digital technologies, including a literature review, a survey of available tools, and a discussion of the uses of AI chatbots in the classroom. It is an excellent point of entry for teachers who want to introduce more digital tools to their students, at any level of education. The second is more theoretical and joins the epistemological threads laid out in the whole book, considering what is digital art history, since “art objects” lack the “intrinsic logical or grammatical features” (p. 12) that make texts and geographical data comparatively easy to digitize. It argues that an art object can be conceived of as a series of annotations, and thus modelled as a collection of “nodes”, each capturing a specific aspect of the art object. Linked Open Data methodology thus becomes a heuristic tool to “think computationally” about art history.
Evolving Perspectives in Digital Classics succeeds in what it sets out to do: many chapters are good first reads for novices; all of them are clear on their methodological premises and contributions to the field; and all classicists, whether trained in Digital Humanities or not, will delight in reading about so many innovative computational approaches.
Authors and titles
Part I. Annotating and Describing Ancient Narratives and Entities
- Structuring the Sights and Stories of Pausanias with Wikidata – John Brady Kiesling, Anna Foka, Kyriaki Konstantinidou, Linda Talatas and Elton Barker
- SLaVEgents: Digital Prosopography of Enslaved Persons in Western Eurasia and North Africa (1000 BCE–300 CE) – Kyriaki Konstantinidou, Elton Barker and Kostas Vlassopoulos
Part II. Ancient Spaces
- The Data of Mythic Spaces – Greta Hawes and R. Scott Smith
- From the Pillars of Heracles to Ecbatana: Digital Representation of Ancient Travel Narratives – Chiara Palladino and Cian Colgan
Part III. Computational and AI-Driven Literary Analysis
- Artificial Intelligence for Classical Literary Texts – Eleni Bozia and Wavid Bowman, David Chong, Srija Dey, Guhan Gnanam, Jacob Hoppenstedt, Aarushi Jain, Mackenzie Morrison, Connor Munjed, Neema Owji, Shae Robinson, Gillian Rodgers, Trey Slaten, Austin Stein, Nitai Stevens and Jordan Wu
- Stylometry for Latin Literary Criticism – Thomas J. Bolt, Karen Bui, Pramit Chaudhuri and Joseph P. Dexter
Part IV. Philological Methods in Digital Text Analysis
- Pretrained Word Vectors for Latin Philology – Patrick J. Burns
- New Methods for Old Worlds Deciphering Ancient Scripts: Integrated, State-of-the-Art Approaches – Roberta Ravanelli, Michele Corazza, Lorenzo Lastilla, Fabio Tamburini and Silvia Ferrara
Part V. Innovative Applications in Culture and Education
- New Technologies for Learning and Teaching Ancient Greek and Latin – Federico Aurora
- Digital Art History, Digital Humanities, and Digital Classics: Finding the Missing Link – Stuart Dunn
Notes