A decade of the Greek Anthology project

Achievements, successes, challenges

Mathilde Verstraete, Marcello Vitali-Rosati

Napoli - 16 Aprile 2024

Horizons de la philologie numérique

Χαῖρε

Welcome

Bienvenue

Benvenuto

The origin of a project…

Looking for an epigram

Φείδῃ παρθενίης. Καὶ τί πλέον; οὐ γὰρ ἐς ᾃδην

ἐλθοῦς᾽ εὑρήσεις τὸν φιλέοντα, κόρη.

Έν ζωοῖσι τὰ τερπνὰ τὰ Κύπριδος· ἐν δ᾽ Ἀχέροντι

ὀστέα καὶ σποδιὴ, παρθένε, κεισόμεθα.

Looking for an epigram

Where is it? Who wrote it?

Looking for an epigram

Φείδῃ παρθενίης. Καὶ τί πλέον; οὐ γὰρ ἐς ᾃδην

ἐλθοῦς᾽ εὑρήσεις τὸν φιλέοντα, κόρη.

Έν ζωοῖσι τὰ τερπνὰ τὰ Κύπριδος· ἐν δ᾽ Ἀχέροντι

ὀστέα καὶ σποδιὴ, παρθένε, κεισόμεθα.

The problem

How to propose a digital editorial model capable of embodying a literary corpus such as the Greek Anthology?

A first platform: Spip

The API

Restructuring data

A new platform (django)

The Greek Anthology project

Collaborative edition and translation of the Greek Anthology anthologiagraeca.org containing:

greek text in multiple editions
translations
scholia
alignment with the manuscript
authors, keywords, places, comments, references…

Goals:

a hub for the Greek Anthology
present the Anthology to a wide audience

The Greek Anthology project

Many actors:

Elsa Bouchard
Christian Raschle
Servanne Monjour
Annalisa Divincenzo
Margot Mellet
Mathilde Verstraete
Luiz Capelo
many other students…

Arthur Juchereau
Tim Guicherd
David Larlet
Sara Rubio
Perseus
Perseids
Library of Heidelberg
Liceo Classico Cagnazzi (Altamura)

What is an epigram?

An abstract entity which has

texts
keywords
manuscript’s images
author(s)
scholia
…

The first question that the manipulation of the data forced us to ask ourselves was: what is an epigram? Associeted with an epigram there were Greek texts - several established texts - translations of these texts, images of the codex palatinus 23, scholies, keywords in several languages… The epigram could not therefore be a text. The epigram had to have texts.

The epigram had to be an abstract entity which had several things: texts in several languages, keywords, authors. etc.

This was also in line with our wish to make an edition for a large public: it was not a question of searching for the truth of the text, but rather of highlighting all the resonances of the anthological topoi.

After the development of the first api, we were able to set up several collaborations: several editors (not only academic, but also amateur, like high school students) started to fill in information: adding texts, translations (existing or original), comments, notes, keywords…

We continued to manipulate this data to visualize and disseminate it: a tweetbot, a site where you could see the epigrams on a map etc.

This programming activity had several effects: a) an increase in computer skills; b) a better understanding of the model and its limitations (what was missing from our data) c) a better knowledge of the data we had produced; and d) a finer analysis of the needs that would have been the basis for the development of a new data model and a new platform (the current one).

Data manipulation and evolution of epistemological models

Semantic Web Alignments

Intelligence artificielle littéraire

HTR

Our trained model: https://zenodo.org/records/10912498

So we started to manipulate the data for research purposes in particular with a large use of Jupyter Notebooks: they allowed us to accurately display the data, to understand the structures and relationships of our model, and to raise major epistemological and methodological questions, sometimes from a data item that was an exception.

The need to disambiguate certain data led us, in addition, to use Jupyter Notebooks to explore other databases and thus try to align our objects with unique identifiers. This is the work we did on the authors: we first retrieved the TLG identifiers of the authors from the XML version of Perseus. From the TLG identifier, we were able to search for the corresponding Wikidata identifiers (when Wikidata presented the TLG identifier and when the author existed in the Wikidata database). We then aligned the authors in our database with the TLGs and the Wikidata identifiers. The work was partly done with algorithms and then verified by hand.

A recent evolution of the project consists in developing algorithms to try to better understand the notoin of variation. We mostly experiment with word embedding approches. The idea is not to find epigrams that are variations of other epigrams, because we already know the variations in the anthology. It is rather to see if an algorithm that manages to find the variations we want can be considered a formal definition of the concept of variation itself