TMI: A Curatorial Approach to Finding Data in a Literary Corpus

morangi3 — Tue, 16 Jan 2018 16:10:37 +0000

Principal Investigator: Brad Rittenhouse
Project Team: Sudeep Agarwal, Taha Merghani, Madison McRoy, Nate Knauf, Sidharth Potdar, and Kevin Kusuma

Informationally-dense literature, sometimes referred to as “encyclopedic narrative,” has often been prized by scholars, and afforded a prestigious place in the literary canon. However, these works and the prestige that comes with them tend to be overwhelmingly male: books like Thomas Pynchon’s Gravity’s Rainbow and Herman Melville’s Moby Dick, for instance, assemble knowledge on topics like ballistics and whaling. This project explores methods for agnostically identifying instances of information aggregation across a literary corpus, sidestepping human biases that overvalue data from mathematics and the hard sciences.

Working on the Wright American Fiction corpus, which includes nearly every work of American fiction written between 1850 and 1875, we have developed a curatorial process that points to specific passages where material information—the people, products, and print that proliferated during this period of American history—accretes. Manifested as noun density, this measure allows us to quantify literary data at a suitable level of specificity: it allows us to find writers who are struggling to represent their newly dense material reality aesthetically, but stops short of proscribing specific types of material data. Writers who catalogue whales, like Melville, are counted equally to writers who may concern themselves more with, say, household items.

Moving forward, we hope to refine our algorithm to detect not just the presence of information, but its absence. African-American writers, for instance, often struggled with assembling hereditary information, which was often kept from them, or in representing experiences too traumatic to be told. We are also working on interactive frameworks for visualizing our data and allowing others to explore it.

morangi3 Tue, 01/16/2018 - 11:10

Project Year

2017-18

Project Leads

Brad Rittenhouse

Students

Sudeep Agarwal

Taha Merghani

Madison McRoy

Genre Tags

Text Analysis

Machine Learning

Topic Modeling

Digital Humanities

Project Site

TMI Project Site

Contact Email

bcrittenhouse@gatech.edu

TOME: Interactive TOpic Model and MEtadata Visualization

morangi3 — Sun, 29 Oct 2017 20:06:17 +0000

TOME: Interactive TOpic Model and MEtadata Visualization

TOME is a tool to support the interactive exploration and visualization of text-based archives, supported by a Digital Humanities Startup Grant from the National Endowment for the Humanities (Lauren Klein and Jacob Eisenstein, co-PIs). Drawing upon the technique of topic modeling—a computational method for identifying themes that recur across a collection—our tool allows humanities scholars to trace the evolution and circulation of these themes across printing networks and over time.

Publications related to this project include:

Klein, L., J. Eisenstein, and I. Sun. “Exploratory Thematic Analysis for Digitized Archival Collections.” Digital Scholarship in the Humanities 30.1 (2015).
Eisenstein, J., I. Sun and L. Klein. “Exploratory Text Analysis for Large Document Archives.’ Proceedings of Digital Humanities 2014. Hamburg: Univ. of Hamburg, 2014.
Eisenstein, J. and L. Klein. “Reading Thomas Jefferson with TopicViz: Towards a Thematic Method for Exploring Large Cultural Archives.” Proceedings of the Implementing New Knowledge Environments (INKE) Annual Meeting 2012. Vancouver: Scholarly and Research Communication, 2013.
Presentations related to this project include:
“Developing and Sustaining Collaborative Research.” Roundtable. Modern Language Association, Austin, TX, January 2016.
“The Carework and Codework of Nineteenth-Century Abolitionist Newspapers.” The Digital Antiquarian, American Antiquarian Society, May 2015.
“Beyond the Digital Surrogate: Discovery and Analysis of Digital Collections.” Roundtable. Digital Library Federation Forum, Atlanta, GA, October 2014.
“The Best-Laid Schemes: Reflections on Three Years of the NEH ODH Data Management Plan Requirement.” Roundtable. Digital Library Federation Forum, Atlanta, GA, October 2014.
“Exploratory Thematic Analysis for Historical Newspaper Archives.” Institute for Quantitative Theory and Methods. Emory University, April 2015. Also presented at Digital Humanities, Hamburg, Germany, July 2014.
“Towards a Thematic Method for Exploring Large Cultural Archives,” with Jacob Eisenstein. Research Foundations for Understanding Books and Reading in the Digital Age, Havana, Cuba, December 2012.