Prof. Lippincott presents survey of recent published scholarship and fledgling investigations at DSAI colloquium

Prof. Lippincott, director of the CDH, presented an overview of recently-published research (Slide deck) to a mixed audience of computational and humanistic faculty. The papers, to be published in the proceedings of the flagship conference for the computational humanities (LaTeCH@EACL), include insights into:

• Effects of the early Christian church on the semantics of Classical Latin, revealing two opposed distributional shifts as the early Church sought to radically adapt specific vocabulary while also consolidating into a standard orthodoxy.
• Literary representation of demographics using orthographic variation in 19th century literature and its relationship to "in-the-wild" linguistic variation, as represented in LLMs.
• Discovery and structural analysis of the historical record from an invisible language community of Armeno-Turkish residents of the late Ottoman Empire.

The colloquium ended with a brief survey of nascent research questions:

• Translating literature requires particular attention to the many ways that one text can refer to another, collectively known as "inter-textuality". How well do machine translation models preserve these relationships, what can be done to improve this, and how can the measured relationships add to our understanding of literary theory and history?
• Poetic meter, or the patterns of stressed syllables, is typically described in a standard vocabulary ("iamb", "foot", "caesura", etc). To what extent are these natural concepts? By training a model that has minimal assumptions but can expand to capture patterns in the data, comparison can be made between human intuitions and concepts derived de novo from the materials.

The CDH pursues computational approaches to domains, questions, and debates from the traditional humanities, such as English, History, and Modern Languages, drawing heavily on its affiliations with machine learning and natural language processing in Computer Science. "The hallmark of our research is the ability to move productively between computation and humanistic interpretation", says Lippincott. "This relies on scholars with a sophisticated understanding of both, and these papers reflect our early success at building that cohort."

Mentioned in this article:

• Tom Lippincott
• Disentangling mixed use of script and language in the Ottoman Empire
• Literary representations of orthographic variation in the Nineteenth Century
• Non-parametric modeling of poetic grammar: regularity and innovation in Chaucer
• Characteristic shifts in the syntax and semantics of Latin: broadening Auerbach's hypothesis-space