I went to the Voltaire Foundation and Wolfson Digital Cluster seminar on Machine Learning and the Experience of Allusion: Experiments in Classical and Eighteenth-Century Poetry yesterday by James Gawley. Beginning with the changing reader of Voltaire, from Voltaire himself to a Mechanical Turk, and contextualising the work within intertextuality as shared subject, stylistic similarity and direct allusions.
Using Mark Edwards’s epic type scenes (which I think come from Type-Scenes and Homeric Hospitality), he outlined the challenges in identifying the scenes between the Iliad and the Aenead and using clusters to remove extraneous data points to be able to read the results. The issues of synecdoche appeared for search, such as ‘death’ and ‘breathed their last’. Using the Tesserae tool and known dictionaries for translation, the machine learning process found a decent percentage of known allusions but was able to see new ones. The process requires some reading to check it as it can be over zealous and it does not handle contexts such as the 18 century French education or the question of how many books of Virgil has Voltaire read?
As a talk about what might be considered data-intensive science, the seminar was enlightening. It was great to hear someone using and thinking about this from both Digital Humanities and Data Science perspectives.
Updated to add link to Tesserae