Measuring Philosophy in the First Thousand Years of Greek Literature

Résumé

In this pilot study, the author applied LDA topic modelling to train a machine to automatically identify philosophical passages in a corpus (produced by the Open Greek and Latin group and the Perseus Digital Library) representing the preponderance of extant works of the first thousand years of Greek literature. The author utilises qualitative data analysis, document vectors produced through topic modelling, and Classics domain knowledge to distil three numeric scores for philosophical text in Ancient Greek. One measures “good and virtue”, while the second score measures “scientific enquiry”, and the third combines the two to measure “philosophicalness”. The scores are applicable to passages, works, and workgroups within the corpus and performed well in identifying works by philosophers and  passages by non-philosophers containing  philosophical components. Thus, this method represents a widely applicable, scalable, unbiased way to find research-relevant passages in a corpus that is too large to be read in its entirety.