Oct 31 2012
Patterns Taken for Wonder:
A Computational History of Modernism
Hoyt Long, Tom McEnaney, and Richard Jean So
Patterns Taken for Wonder was a project that had two main goals. First, to provide a new history of literary modernism through the lens of computational and data methods from the social sciences, such as machine learning and social network analysis. Beginning in 2013, we assembled a large database of over 50,000 modernist poems, thousands of novels, and hundreds of “little magazines” from the United States, England, Latin America, China, and Japan as part of our Global Literary Networks project. We sought to leverage this corpus to explore new claims about the history of modernism at scale, searching for patterns and structures otherwise invisible to the human eye. We aimed to combine these interpretations with traditional close reading and historicist methods.
Second, the project tried to sketch a history of the idea of computation and algorithmic reading within modernism itself. That is, we see literary modernism as anticipating many contemporary ideas about automated text analysis we find in today’s data sciences. For example, we see this in Marinetti’s interest in language as a mere “system” composed of “atomized” pieces that can be constantly recombined, which resonates with the “bag of words” approach popular in Natural Language Processing. We also find this in the work of prominent modernist scholars, such as Hugh Kenner, who were drawn to the use of computation to process and find patterns within modernist texts. In sum, we discern in modernism an important prehistory to our current “big data” moment in which the past can be read forward into the present, and vice versa, the two mutually illuminating.
Our project sought to bring these two strands together to model a new form of literary criticism. It is a model that attempts to integrate machine and human forms of interpretation, taking seriously the way that a machine understands a text, and putting that into dialogue with conventional human/humanist epistemologies of literature and art. We called this method of reading “literary pattern recognition.” We believe that the intelligent and critical synthesis of machine and human modes of explication can reveal new patterns of meaning both within individual texts, as well as aggregrates of texts at scale.
The title of the project was a nod to Franco Moretti’s important Signs Taken for Wonder, which explored literary texts as not merely singular examples of expression, but as literary systems that are tokens of greater social and political realities. This book was important in advancing a modern sociological approach to literary study. We saw our own work as extending this project: to reread the modernist text as in part constituted by and animated by algorithmic patterns of meaning not incommensurable with current machine models of interpretation; and to see texts as not only sociological tokens, but also, as contributing to and in part shaped by vast patterns of culture. Most broadly, we aimed not for a simple “application” of big data methods for literary criticism, but an integration of humanist and scientific approaches to the text in order to model a new form of cultural criticism. And with this model, we sought to write a new history of modernism, larger and more empirical than previously seen. Importantly, our case studies also respond to the recent transnational turn in modernism studies by working through canonical and non-canonical, Western and non-Western, examples.
Parts of this project have appeared in several journals. “Literary Pattern Recognition: Modernism Between Close Reading and Machine Learning” was published in Critical Inquiry in Winter, 2016. “Turbulent Flow: A Computational Model of World Literature” is forthcoming at MLQ in Fall, 2016. Please feel free to contact Hoyt Long or Richard Jean So for preprints. We are always eager for feedback and comments.