Three Forms of Historical Description. Selecting Coverage from the New York Times over Many Decades in Pursuit of Continuity.

Alix Rule, New York University (NYU)

This paper argues that computational text analysis can help historical researchers develop more rigorous definitions of continuity. The task of automating the selection of newspaper coverage consistently over the longue duree, for example, reveals that continuous descriptions of social context over time take different forms. I draw on my work on document selection with J-P Cointet: Relying on neural embedding techniques, we model the organization of the social world visible in the coverage of The New York Times 1900-present. We can navigate this model historically, compiling collections of ledes (events) that represent a given social context from distinct, but equally consistent standpoints, over many decades. Using the case of the state mental hospital at midcentury to illustrate: we may consider how the apparent context of the state hospital changed, what in the social world replaced the state hospital over time, or how the context that the hospital occupied at midcentury itself evolved. By abstracting these document collections, we achieve three forms of historical (ie diachronic) description, each with precedent in the work of historians.

No extended abstract or paper available

 Presented in Session 34. Big Data and Its Discontents: Assumptions about Reading History in the Automated Analysis of Texts