In this blog, I will discuss three paper about the graph analysis of main characters and events in the novels. It is especially useful to visualize the novels as network diagrams if we want to clarify the relationship between main characters and events.

The first paper introduce a new technique (character network analysis) to model and distinguish different genre. They apply this method to compare the National Tales and Domestic Novels of Maria Edgeworth. After collecting this raw data in the novel, they create the adjacency matrices for each chapter. In the network graph, every node represent a different character and every edge was drawn when two corresponding characters get in touch with each other. And the weight of a edge is equal to the number of chapters in which they speak to each other. By using Gephi package, they plot the graph and calculate three different measurements: degree centrality, betweenness centrality and strength to centrality. The analysis result reveals magnificent difference between ‘National Tales’ and ‘Domestic Novels’. In a word, we can grasp an important similarity and difference between different novels if we visualize the novels as network graphs.

The second paper illustrates how to use network analysis methods to categorize the novels. They use centrality measure such as degree and eigenvector to analyze the character network. They construct modelling the network using 20 novels in Les Rougon-Macquart. After collecting the data, twenty small tables is created. They use one set of the nodes to represent the characters and the other set composed the pages. If the character appear in this page or sentence, there exists a link edge between them. Then we can obtain character’s network by projections. By observing the density of a network, they can measure the associativity between each character. From studying centralization and coreness, they can categorize and compare the character network. By iteration, they can analyze the large character networks.

The third post show a preliminary answer about the question ‘how do you do with a million readers’. As the explosion of world literature, equally explosion of readers’ comment appear. They use sixteen fiction as the target data. Their initial goal is to develop a summary network about the relationship between the main characters and events. Also they analyze the reader comments using two metrics: completeness and accuracy. They chose the ‘SparkNotes’ as the gold standard summaries. They face several challenges about how to discover the main dramatis personae, how to discover the important event related with main personae, and how to visualize the relationship clearly. To solve these problems, they eliminate the ‘noisy’ comments in the preprocessing stage. Then use three approach that LDA is especially successful among them to discover dramatis personae in the fiction. To discover the relationship, they extra the verbs and use POS tagging between entites. They create lists of nous containing names, location, objects and concepts to evaluate using ‘SparkNotes’ summaries. Also they developed somes multicolor graphs to visualize the strength of the relation and create some other graphs to compare different comments for different types of reader.

As these three abstract shows, network graph will be widely used to analyze characters in the book for its clearly representation and better statistical. Accurately, we can easily transform the graph to matrix in the computation. By analyze the character graph, we can grasp the similarity, the difference, the relationship between characters. In the first article, they use the network diagram to compare the character in different type (National Tales and Domestic Novels) of novels written by Maria Edgeworth. While the second article reply on method to track down major and minor characters using degree and engienvector as the measurement. The third article focused on the comments of the reviews. Differently, they use the comments to model the character network and compare the different comments by different classes of reviews.


Michael Gregory Falk. Modelling Genre Using Character Networks: The National Tales and Domestic Novels of Maria Edgeworth

Yannick Rochat. Character Network Analysis of Émile Zola’s Les Rougon-Macquart.

Roja Bandari, Timothy Roland Tangherlini, Vwani Roychowdhury. What Do You Do With A Million Readers?