Patrick J. Burns

Research Associate at Harvard Human Evolutionary Biology | Formerly Quantitative Criticism Lab, ISAW Library | Fordham PhD, Classics | CLTK contributor

Error, Scale, and the Computational Redefinition of Philology

Abstract for Digital Humanities Beyond Modern English 2022 conference
April 7, 2022. Austin, TX.


Philology is a discipline defined by the rejection of error. As one critic remarks, it is a field aimed at “reconstructing what the author must have written rather than enshrining or embalming the errors transmitted.” (Zetzel 2015) Because of this disciplinary disposition, the use of computational tools in philology presents something of a methodological challenge: on the one hand, these tools offer us the ability to work with historical-language text at unprecedented scale, while on the other hand these tools also introduce various sorts of errors—from OCR mistranscriptions to faulty annotations—and the practice of working at this scale makes comprehensive review and correction difficult if not impossible. Charlotte Schubert (Schubert 2019) has recently advocated for a “constructive culture of error” in digital classics, that is the adoption of practices that diligently and systematically document errors that arise in the course of research in order to defend this work against criticism. In this chapter, I push the idea of “constructive” error one step further, choosing not only to defend computational philology against criticism but to highlight positively the advantages that these tools bring to the study of historical-language texts even when the introduction of error is unavoidable. Computational approaches have recently shown promise in tasks as varied as epigraphic reconstruction (Assael et al. 2022) and the study of manuscript loss (Kestemont et al. 2022) not because they advance new philological certainties but because they responsibly model philological probabilities. I argue here that philology needs a redefinition to reflect the massive expansion of research questions made possible through computational approaches, a definition less concerned with the eradication of error and more concerned with leveraging the affordances of effective, if at times flawed, tools and models.

Works Cited

rss facebook twitter github youtube mail spotify instagram linkedin google google-plus pinterest medium vimeo stackoverflow reddit quora hcommons