data Directory

The dataset is based on Folger Shakespeare editions encoded as XML documents, which are freely available on their website: https://www.folger.edu/explore/shakespeares-works/download.1 These XML editions are stored in the data directory of the GitHub repository. The XML documents represent digital transcriptions of Shakespeare’s plays and are used to encode documents in a format that is both human-readable and machine-readable. The Folger Shakespeare editions in XML format provide detailed textual data, including character dialogue, stage directions, and other critical annotations. Importantly, these XML editions offer accurate and machine-readable information on the entrance and exit of characters, which is crucial for analyzing character interactions and dynamics. These files are essential for the dataset as they offer a standardized, rich textual source that can be parsed and analyzed to extract character interactions and other relevant information for the project. By utilizing these XML files, the dataset ensures accuracy and consistency in representing Shakespeare’s original texts.

output_onstage Directory

The output_onstage directory contains files that document the onstage interactions of characters across Shakespeare’s plays. These files in .csv (Comma-Separated Values) format are created using the main.py script, which processes XML files from the Folger Shakespeare Library to extract and quantify character interactions. The script analyzes each line to identify which characters are present together on stage and then constructs matrices representing these interactions. In these matrices, the rows correspond to characters and the columns represent Folger Through Line Numbers (FTLN). The FTLN columns indicate specific points in the narrative, effectively mapping the temporal sequence of the play. The cell values within these matrices denote the presence or absence of characters at particular narrative points, with 0 representing absence and 1 representing presence.

output_onstage_heatmap Directory

The output_onstage_heatmap directory contains heatmap visualizations that depict the onstage co-presence of characters. These heatmaps are generated from the data in the output_onstage matrices, providing a visual representation of how often characters appear together in scenes. The heatmaps facilitate a quick and intuitive understanding of character interactions and dynamics, making it easier to identify clusters of characters based on their onstage or offstage status. HTML files in the directory can be opened using any standard web browser, allowing for an interactive exploration of the visualizations. Additionally, the .png files in the directory are non-interactive versions of the .html files with fewer details. These static images can be easily viewed without the need for a web browser.

output_onstage_matrix Directory

The output_onstage_matrix directory contains .csv files that provide interaction matrices derived from the matrices in the output_onstage directory. Each file contains a matrix where both the rows and columns correspond to the characters in a play. The cells in these matrices quantify the number of lines in which each pair of characters appears together, allowing for a further analysis of character co-presence and interactions throughout the play. These matrices are essential for conducting network analysis and understanding the structural relationships between characters in Shakespeare’s plays.

Additionally, the directory includes .html files, along with .png files with fewer details, that visualize these interaction matrices using the NetworkX library. NetworkX is a Python library used for the creation, manipulation, and study of complex networks of nodes and edges.2 The visualizations employ a spring layout, which simulates a physical system where nodes repel each other while edges act as springs pulling related nodes together.3 This layout helps in creating aesthetically pleasing and informative representations of the network structure. In the visualizations, the shape of the nodes represents different groups of characters identified using the K-means clustering method with k = 2, while the color of the nodes indicates different groups of characters determined by the Louvain method.4 The K-means method partitions the characters into two clusters based on their interactions, while the Louvain method, a community detection algorithm, identifies clusters of characters that are more densely connected internally than with the rest of the network. This combination of methods provides a comprehensive view of character dynamics and their social structure within the plays.

output_onstage_matrix_modified Directory

The output_onstage_matrix_modified directory contains modified versions of the interaction matrices found in the output_onstage_matrix directory. These modified matrices have been refined to remove characters who are mostly isolated with almost no interactions with other characters, such as prologue, epilogue, and citizens 1, 2, and 3 in Richard III. These outliers can interfere with community detection methods by appearing as standalone communities, making it difficult to understand the clustering patterns of the majority of characters. By removing these outliers, the matrices in this directory provide a clearer and more useful representation of the main character interactions, thereby facilitating more effective network analysis and community detection.

output_exchange Directory

The output_exchange directory contains matrices that document the amount of dialogue exchanged between characters. Unlike the output_onstage_matrix files, which focus on the presence of characters on stage together, the output_exchange matrices quantify the number of words spoken between each pair of characters. The main.py script processes the XML files in the data directory to calculate these exchanges. It assigns speakers and listeners for each line of dialogue by parsing the text and identifying who is speaking. The assumption made is that all characters on stage are listeners while a speaker gives his or her lines. In these matrices, the columns represent the speakers, while the rows represent the listeners. The cell values indicate the total number of words spoken by one character to another, offering a detailed view of verbal interactions. These matrices are essential for understanding the communication dynamics within the plays, highlighting which characters interact the most through dialogue. By analyzing these dialogue exchanges, researchers can gain insights into the relationships and influence of characters based on their verbal interactions. The data in the output_exchange matrices provide a complementary perspective to the onstage presence data, offering a more nuanced view of character interactions within Shakespeare’s plays.

output_exchange_heatmap Directory

The output_exchange_heatmap directory contains heatmap visualizations of the dialogue exchanges between characters. These heatmaps are generated from the data in the output_exchange matrices and provide a visual representation of the intensity and frequency of verbal interactions. The heatmaps help to quickly identify which characters engage most frequently in dialogue, making it easier to discern communication patterns and key relationships within the plays.

output_exchange_modified Directory

The output_exchange_modified directory contains modified versions of the interaction matrices found in the output_exchange directory. These files have been refined similarly to those in the output_onstage_matrix_modified directory, removing characters who are mostly isolated and have minimal interactions with other characters.

output_scores Directory

The output_scores directory contains files that calculate various network centrality measures and clustering indices for both exchange and onstage interaction matrices. Specifically, the degree centralization scores, betweenness centrality scores, closeness centrality scores, eigenvector centrality scores, Dunn Index, and silhouette scores are computed for the networks derived from the character interactions. Degree centralization measures how centralized the network is by evaluating the variance of node degrees, with the formula summing the differences between the maximum degree and each node’s degree, normalized by the theoretical maximum. Betweenness centralization identifies nodes that act as bridges by measuring the number of shortest paths passing through each node, summing the differences between the maximum betweenness centrality and each node’s betweenness, and normalizing this sum. Closeness centralization assesses how quickly information spreads from a given node to others by calculating the average shortest path length. The centralization score sums the differences between the maximum closeness centrality and each node’s closeness, then normalizes the total. Eigenvector centralization evaluates the influence of nodes based on their connections to other high-scoring nodes, summing the differences between the maximum eigenvector centrality and each node’s eigenvector centrality, and normalizing the total. The Dunn Index and silhouette scores are used to evaluate the quality of clusters formed by the character interactions. The Dunn Index is calculated as the ratio of the minimum inter-cluster distance to the maximum intra-cluster distance, highlighting the compactness and separation of clusters.

These scores are visualized using bar graphs and box plots, generated by the plot_scores function in the main.py script. In the bar graphs, the x-axis represents the different plays, ordered chronologically according to the conjectured year of composition as per the New Oxford Shakespeare edition to maintain consistency and ease of comparison across the visualizations.5 The y-axis shows the respective centrality or clustering score, allowing for a comparative view of how different plays rank in terms of their network metrics. Different colors are used for the bars to distinguish between genres of the plays: comedies are colored light blue, tragedies are colored slate gray, histories are colored red, and plays with an unidentified genre are colored light gray. This color-coding facilitates an immediate visual distinction between different genres, enhancing the interpretability of the graphs. The box plots provide additional insights into the distribution and variability of these scores across all plays. For each centrality measure and clustering index, the box plots depict the median, quartiles, and potential outliers, offering a more detailed statistical summary. These visualizations facilitate a comprehensive understanding of the structural and communicative dynamics in Shakespeare’s plays, highlighting variations in character interactions and network properties across different works.

Python Code: main.py

The main.py script in the repository (see fig. 1) serves as the primary engine for processing and analyzing character interactions within Shakespeare’s plays. It begins by importing necessary libraries such as scikit-learn, Community API, and NetworkX, and defining key functions for parsing XML files from the Folger Shakespeare Library. The script reads these XML files to extract detailed information about character dialogues and stage directions. By parsing this data, main.py generates various data structures, including onstage presence matrices and interaction matrices, which are then stored in the appropriate directories. These matrices quantify character co-presence and dialogue exchanges, providing a foundation for further analysis. The script also includes functions to compute network centrality measures and clustering indices, allowing for a detailed exploration of the character networks within the plays.

Figure 1.

Snippet of code in main.py