Digging into the Enlightenment with Hypergraph Queries

Update: This work was published in Volume 19 of IEEE Transactions on Visualization and Computer Graphics! Full text here: Visual Analysis of Higher-Order Conjunctive Relationships in Multidimensional Data Using a Hypergraph Query System

I had the pleasure of doing my thesis research as part of the Digging into the Enlightenment collaboration. Digging into the Enlightenment “seeks to discover that lost continent through an exploration of empirical data gleaned from correspondences, publications, and travel records, combined with the interpretive expertise of historians and literary scholars. Mapping the Republic of Letters is an opportunity for a unique collaboration between the humanities and sciences to produce a model of a real world network, using rich and diverse examples from this historical material.”

The project was a collaboration between humanities researchers at Stanford University, data archivists at Oxford, and data visualizers at the University of Oklahoma (that’s us!). The data we worked with is the exceptional Electronic Enlightenment, which is an online collection of letters from the mid 17th to early 19th centuries. The EE data set includes over 58,000 letters sent during the Enlightenment, in addition to information about the more than 7000 letter senders and recipients. The database contains a variety of details about the letters, including the primary language the letter is written in; the letter’s source and destination city, state, and country; the date the letter was sent; the nationality, birth place and date, death place and date, and occupations of the authors and recipients; and the age of the author and recipient at the time a letter was sent, to name most.

Our role in the project was to develop tools to help the humanities researchers explore patterns in this rich data set. My thesis work focused primarily on allowing the researchers to ask complex questions* of the data.  To do this, I developed a visual query language using hypergraphs and incorporated it into a visual analysis tool called Candid. Candid was developed using Improvise, which is an open-source system for building interactive data visualizations, to which I added the ability to process hypergraph queries. Query results can be displayed in a variety of ways, but in this case I chose an attribute relationship graph. Nodes in an attribute relationship graph represent values of attributes. Edges between two nodes represent letters that have both of those attributes. The thicker the edge, the more letters there are that have both of those attributes.

Below, you’ll find an example of Candid in action! In this example, the question being posed is about the political activity of clergy in a portion of Europe. We want to know which clergymen are writing to which politicians, and where both parties are located at the time of writing. Our answer is that a clergyman in France, Pierre Hugonet, was writing to Joseph Marie Balleydier, a politician in France, while Frederic Guillaume de Montmollin, a clergyman in Switzerland, was writing to two politicians in Switzerland, Fracois Henri d’ Ivernois and Jacques Francois Deluc.  The video shows how the hypergraph query is constructed. For best results, watch large and in HD.

*I define “complex questions” as queries involving higher-order conjunctive relationships. Most existing techniques can only ask questions about binary conjunctive relationships. If these details are of interest to you, I am happy to supply you with some light thesis reading.