Quote

“Discovery is a very human activity. . . it shouldn’t be restricted to scientists.”

– John Yin, Professor of Chemical and Biological Engineering and Theme Leader, Systems Biology at UW-Madison & the WID

Advertisements

Digital Humanities at UW-Madison

The Digital Humanities (DH) is an emerging scholarly approach to the questions we have been investigating in literature (and other areas of study in the humanities such as sociology, art, and psychology) that combines the scientific data sets collected from literature with a literary analysis and interpretation. It is an intersection point between two disciplines with seemingly different origins. Through the use of statistical and computation sciences, we are able to think about ideas in a new way, and explore the way that this data reveals new patterns and concepts that begin to answer some of the unanswerable and difficult questions and challenges faced by literary scholars. The UW-Madison Humanities Research Bridge (explained in the post below) has many such projects, and specifically the ScripThreads Tool created by Eric Hoyt, Kevin Ponto and Carrie Roy explores character interactions and presence throughout a narrative in a visual and interactive manner.

The Digital Humanities changes the way we read and experience text, and specifically narrative. As a reader, narrative is experienced in, and through time, both at the specific point you are reading in time, and time throughout the narrative. DH offers a new perspective and approach to narrative in that we can engage in “distant reading”, which involves experiencing and analyzing the narrative, novel, play, etc. as one, whole entity.

Moretti's character network of Hamlet

Moretti’s character network of Hamlet

This can be accomplished through Franco Moretti’s character networks and maps (Network Theory and Plot Analysis), which filters the text down to interactions between characters, which pioneered the practice of “distant reading” as a form of contemporary literary analysis as opposed to the more traditional “close reading”, which focuses on smaller, individual portions of the text in relation to a theme or thesis.

DH takes these networks one step further by including a strong visual component to the experience of character networks and maps. Just as Moretti quantified character interactions into edges and nodes between character names, the DH similarly quantifies a wide range of characteristics (age, gender, communication, social status, etc.) into data sets and creates interactive tools, methods, and maps to analyze literature in an entirely different way. I will be focusing specifically on Eric Hoyt, Kevin Ponto and Carrie Roy’s ScripThreads Tool, and how it enhances our distant reading literary analysis of texts and screenplays.

The Humanities Research Bridge

The Humanities Research Bridge is a digital research community at UW that works to collaborate with other institutions and individuals to put together workshops, provide consultations, and spotlight support projects done by graduate students, scholars, and the UW community. The HRB has been working to collaborate with a network of cutting edge digital research and scholars that are interested in humanities projects to put together a pool of resources and tools for digital humanities scholars to utilize, challenge, and engage with. It works with the aid of NEH grants as well as MacArthur Foundation Digital Media & Learning grants. 

Many of the current projects that HRB is working on explore narrative networks, character analysis, textualization, and archival work. The project we will be focusing on is the ScripThreads Tool, by Eric Hoyt, Kevin Ponto and Carrie Roy.

Video

ScripThreads Tool Video

Carrie Roy created an informational video explaining the ScripThreads Tool (in this video titled the Screenplay Narrative Threads tool) which highlights many of the aspects of the project that I will be discussing and analyzing below. The various graphs displayed in this video are now labeled under different titles, as this video was made before the tool was finalized and completed. The most recent and up-to-date terminology surrounding the ScripThreads Tool are included in my profile of the project.

ScripThreads Profile

The Digital

The ScripThreads (ST) tool is a way to visualize character interactions according to screenplays. The tool visually divides the scenes of the screenplay with different stripes, each stripe indicating a change in scene. The analyzer of the tool is able to interact and manipulate the visual “image” that is created by the tool, scrolling down the entire course of the screenplay, pivoting the image to see various connections, increasing the graph, and zooming into various scenes and sections of the image to see the specific scenes. This ability to zoom in and pivot the image allows the analyzer to participate in both close and distant reading at the same time.

The ST tool works through a force and physics directed algorithm, which is used to visually graph the character interactions within the space of the screenplay. When the characters entered into the algorithm are in the same scene, their respective graph lines connect in the middle, and when not in the scene, the lines will loop out. The data set for the ST tool is specific for screenplays, and any visualizations generated from novels or any other forms of text must be reformatted to the ST tool’s qualifications. The format of a screenplay, which structures the text in specific divisions of scene changes, dialogue and character names and entrances and exits. These specific formatting divisions allow screenplays to be converted to HTML which is then converted by the tool into the visualizations and graphs used to analyze the screenplay. This process is explained by the project designers in a forthcoming article, “The tool is written in C++ and utilizes the QT toolkit for its graphical user interface. . . The tool takes in text and HTML file screenplays as an input, parses these files and generates data for visualization and analysis” (Hoyt, Ponto, Roy; hyperlinks provided by me as definitions of the mechanisms of the tool). The parsing method is used to locate characters in the screenplay and quantify lengths of scenes (by page number, not screen time) and the characters that interact within (Hoyt, Ponto, Roy). This data is then manipulated and generated into various visualizations. 

The ST tool has various ways to manipulate the data, allowing the analyzer to view the data in four different ways, focus on one character’s role, and pivot the image to better see the linear relationships between the characters (a 3-D image). The data can be viewed as force directed graph, an absence graph, a presence graph, and an increasing graph.


 

Force Directed Graph. Used with permission from the authors

Force Directed Graph. [Used with permission from the authors]

The force directed graph uses the data collected to create a 3D network of character threads that relate time and character actions in the scenes of the screenplay. This graph is presented vertically alongside the corresponding scenes in the actual screenplay, which allows the viewer to participate in both close and distant reading at the same time. The character threads have a changing diameter depending on whether or not the character is currently active in the scene (thick vs. thin). As character interact, their corresponding threads connect and intersect (Hoyt, Ponto, Roy). The force directed graphs allow the viewer to analyze the entire screenplay or novel as an entire structured narrative, enhancing our distant reading capabilities. As we can see more than one scene and more than one character acting at parallel times (which is impossible while in the act of actually reading the narrative or close reading the narrative). This ability to see the text as one whole entity allows the viewer to make connections between characters that may never interact directly, but interact in a more thematic or tropic way.

 


Absence Graph. Used with permission from the authors.

Absence Graph. [Used with permission from the authors]

The absence graph loops the character threads away from the center as smooth arcs, with the distance it loops out dependent on the time that they are absent from the scenes. The x-axis here represents whether a character is present or absent, and the distance it loops out from the x-axis is the time they are absent (time here is not linear in one direction, but can be viewed as bot the past (up) and future (down) from the present node at the x-axis. In a forthcoming article, the project designers describe the visualization of the absence graph as being “read like a bus map: characters run parallel routes when they both appear in a scene. When a character is not in a scene, his or her bus route splits off” (Hoyt, Ponto, Roy). This description and the visualization underscores the concept that although a character is no present in a scene, he or she still has a connection to the scene and a relationship with it: his or her absence. Although he or she is not in the scene, he or she is still following his or her own route through time, a parallel narrative that we may not be given insight to. Just because some characters are not present in a certain scene does not mean that they do not exist at that point in time. In the frame of distant reading, we are able to visualize the entire narrative and analyze the importance of the absence of a character in relation to the direction of the narrative, and the interactions between the different characters in relation to his or her absence (i.e. why is it significant that two or more characters never appear in the same scene, or how does a character’s absence influence the interactions between other characters).


 

Presence Graph. Used with permission from the authors

Presence Graph. [Used with permission from the authors]

The presence graph quantifies a character’s presence in a scene by varying the width of his or her character thread (i.e. when a character is present, the thread is thicker and when he or she is not present, the thread is thinner). These threads run along the vertical y-axis with horizontal lines between characters denoting dialogue (Hoyt, Ponto, Roy).


Increase Graph. Used with permission from the authors.

Increase Graph. [Used with permission from the authors]

The increasing graph is a view of the text that doesn’t focus on character networks and network theory, but rather focuses on how the characters individually are active throughout the narrative (relating time to the activity of the characters, allowing the viewer to see which characters were the most present in the narrative as we move through the screenplay or text). Although it doesn’t focus on character interactions, it does place all of the characters’ lines on the same two axes. The increase graph creates strictly linear relationships and focuses more on a 2-D linear visualization of character activity compared to the more interactive and 3-D visualizations created by the other three graphs. The x-axis depicts time, with the rising y-axis depicting whether or not the character is present in the scene (time). As time increases, a character’s thread will either increase if he or she is present in the scene, or remain flat if they are not present (immobile in time, compared to the motion of looping out from the axis as depicted in the absence and force directed graphs). This graph can be used to compare gender roles (i.e. if male or female characters have more activity in the narrative), social status and other social divisions (race, political factions, religion, etc.), and the voice of the narrative (1st person or 3rd person narratives). For example, a 1st person narrative would have a much steeper slope for the protagonist’s activity in the narrative in comparison to the other characters as the entire narrative is told from his or her point of view. Contrastingly, a 3rd person omniscient narrative would have variable amounts of steep (strongly active characters) slopes, as the narration can follow multiple different protagonists and characters throughout the narrative time.


 

The Digital Humanities allow us to change the way we think about ideas, by using data sets and computation approaches to present narratives in a different analytical framework. The ScripThread tool takes Moretti’s network theory to an entirely different level by introducing the interactive and three-dimensional aspects of the graphs and visualizations generated by the data set. This three-dimensionality allows for the interaction and convergence of both close and distant reading simultaneously. These interactive visualizations expose the dynamic relationships between characters over time that are hidden when we are focusing on specific moments in time during close reading analysis. As time is difficult to conceptualize while reading and especially close reading, the exposure to this third dimension found in the ScripThread tool directs us to new connections between characters and the narrative as a whole. These new digital connections give us a fresh approach to narratives, and pave the way for literary analysis grounded in data sources that allow us to answer previously unanswerable questions and asking new and thought provoking questions that challenge readers to think in a new way.

 

 

ScripThreads Profile

The Human

Although the digital tools give us the ability to experience and analyze narratives in a new form and context, these tools cannot stand alone as an analysis of the text. In order to successfully analyze these texts in the Digital Humanities, we require the HUMAN element. As the project designers explain in a forthcoming article:

“A computer can read the same screenplay as you, but it is going to read it differently. The continuities and differences between your perception and the computer’s is a powerful starting point for uncovering storytelling techniques, better understanding cognitive reception, and achieving a fuller understanding of the object being studied” (Hoyt, Ponto, Roy). 

This passage underscores the importance of the relationship between digital tool and human perceiver. Without the two working in conjunction, or in opposition with each other (the perceiver challenging the tool’s production, and the data reciprocally challenging and further complicating the perceiver’s reading of the narrative), the digital humanities research falls flat: no questions are being asked or successfully answered. As the designers underscore: the digital tools are just “starting point[s] for uncovering storytelling techniques”, and the human perceiver is necessary for interpreting and uncovering the questions, connections and relationships in the narrative that are being studied.

            The project designers further empower the reader: “The emotion is within you. Emotion is something you are great at feeling as reader and spectator” (Hoyt Ponto, Roy). This emotion is meaningless to a computer and digital tool, which follows its specific pathways for detecting data, emotion and cognitive intuitions are unperceivable and uninterpretable to a computer. The designers addressed this digital limitations by including more data sets for the program to detect. As the explain: “we also sought to create a narrative profile. . . that accounted for scenes, pacing, and character interactions. Rather than reducing a screenplay simply to statistical aggregates, we wanted to map the way a screenplay unfolds as it moves from page to page” (Hoyt, Ponto, Roy). The creation of this visual and 3-dimensional map and narrative network combats these limitations by expanding the data set available for interpretation, but it still deals entirely with data and requires a human interpreter.

            Although there are great limitations to interpretive research done by a computer, the symbiotic relationship between the digital tool and the human perceiver strengthens research done by the perceiver/reader. This relationship may be seen as symbiotic (both participants dependent on each other for success (in biology seen as survival), but I argue that it takes on an almost soft parasitic quality of literary analysis. The computer’s data cannot survive, or be successful in interpreting and analyzing the text on its own, and as the human perceiver uses this tool as a “host”, it participates in a more successful analysis of the text. This relationship is a backwards parasitic relationship, where the host (human perceiver) is more successful with the parasite (digital tool), while the parasite cannot survive without the host: a mutual parasitism.

            The human aspect of the Digital Humanities is essential to successful and thought provoking literary analysis. As the project designers close their forthcoming article: “ScripThreads is a tool to aid Humanities scholars. This is a tool for narrative analysis and interpretation, not a substitute for screenwriting and criticism” (Hoyt, Ponto, Roy). Although the tools often overshadow the human involvement in the Digital Humanities, the role of human interaction with both the text and manipulation of the digital tool is integral to the process of literary analysis and the future of the Digital Humanities.

Applications & Limitations

The ScripThreads tool has been developed as a tool used specifically for analyzing screenplays, but there is potential for larger texts, novels, and poetry (usually epic poems) to be integrated into the tool. This involves a more human-controlled data selection than the screenplays, as the texts have to be converted into a screenplay format, with chapters being made into scenes, character names being BOLDED, and dialogue formatted in a way that the ScripThreads tool detects it. This text conversion also involves including entrances and exits and translating the text into the proper HTML version.

As the text is filtered down to the framework of a screenplay, we lose the descriptive aspects of the narrative, as well as the setting and some connections between characters. Many subjective decisions have to be made as to what constitutes “dialogue” to be included in the screenplay: how do we quantify interactions between these characters that at many points throughout the narrative may interact in ways other than dialogue in a screenplay format, but by eye contact, touch, or thoughts? Does being in the same “scene” of the book constitute as dialogue or a connection? In order to obtain the analytical visualizations generated by the ScripThreads tool, we lose a lot of material that could also be used for analysis.

These limitations do not completely invalidate the use of the ScripThreads tool for analysis of books and other forms of narratives that are not naturally read in the format of a screenplay. This tool can be used to supplement and enhance our reading of the narrative, and it allows us to participate in distant reading of the text alongside close reading of specific passages.