Networks and the Digital Humanities THATCamp Pittsburgh October

Networks and the Digital Humanities THATCamp Pittsburgh October 5 th and 6 th, 2013 Dr. Elaine Frantz Parsons Associate Professor of History Duquesne University & Dr. Thomas Lombardi Assistant Professor Computing & Information Studies Washington & Jefferson College

From Graph Theory to Networks Vertices Alice People Chuck Edges Friendship Bob

Constructing Networks: Method Chris t Mary Joh n Giotto, Crucifixion, Tempera on wood, ca. 1290 -1300. Source: commons. wikimedia. org

Network of Saints in Images of St. Francis

Analyzing Network Models Centrality Degree Betweenness Weighted Edges Directed Edges Clustering Affiliation Networks Community Detection

Degree Centrality Number of links connected to a node Node A has a degree centrality of 6 Nodes B, C, D, E, F, G: degree of 1 Degree centrality measures the ease with which information can reach a particular node

Betweenness Centrality % of shortest paths running through a node H has a low degree, but high betweenness Cut vertices Bridges Betweenness confers the ability to control the flow of information in a network

Network of Characters in Hamlet Franco Moretti. Network Theory, Plot Analysis. New Left Review. 2011.

Edge Weights Some links between nodes are more important than others Edges can be weighted to record the importance of a connection For example, some characters interact more than others in a novel Each interaction increases the weight of the link between characters

The Rhyme Scheme of The Raven

Edge Direction Vicomte de Valmont Marquis e de Merteuil Letter 44: The Vicomte de Valmont to the Marquise de Merteuil

Dangerous Liaisons: Network of Letters

Clustering Some nodes are situated in extremely dense parts of the network Node E is entirely embedded: all of the node’s neighbors are neighbors Other nodes are in relatively sparse parts of the network None of I’s neighbors are neighbors

Network of Relationships in Florence Padgett & Ansell, 1993 Network structure accounts for Medici rise to power in the 1430 s Brokerage Constraints amongst old families of Florence made them ineffective Clustering

Affiliation Networks often track more than 1 kind of entity Students and their classes These networks are flexible because they can be transformed into a network of students and a network of classes Networks can also have more than one type of link: multi-relational networks

Union County Affiliation Network

Community Detection In large networks, we frequently want to find the dense clusters of nodes These often represent communities of interest These communities often relate to other groups in interesting ways

Communities within William Faucett’s KNeighbors

Hands-on Activities Flash drives contain some of the networks we’ve discussed Perform some basic analysis of a network with Pajek Create your own network model for analysis in Pajek Experiment with networks