Finding New Ways to Feel Violation: EMDA Final Presentation

[This post is a transcript of my final presentation given July 2, 2015 at the conclusion of Early Modern Digital Agendas: Advanced Topics at the Folger Shakespeare Library. The slides are provided as well, but in PDF form, which truncates the original PP slides a bit. I’ve also included several of the more interesting slides in the body of the text itself. The first part of this talk describes the digital work I’ve already done on “Feeling Violation: Tactile Rendering of the Early Modern Blazon” and the second half gives an overview of how EMDA has shifted my thinking, and some ideas for new trajectories based on a new set of tools and questions. Some moments of the transcript are particularly stilted because it was meant to be an outline for the talk. Any feedback is very welcome, but please do not reproduce any content from the presentation, as it is still in the early stages of a much longer process.]


(SLIDE 1) First, I want to thank you all for an amazing two and a half weeks. The conversations we’ve had in this room and my discussions with many of you more one-on-one have been a tremendous help in honing my ideas and learning more about the overwhelming range of tools available to help me execute them.

My plan for this presentation is to briefly introduce the project I came to EMDA with. What my major research questions were and how I’ve gone about answering them thus far. And then I’ll spend most of my time on how EMDA has shifted my research questions and what my plan is moving forward.

The project that I’ve been working on for about a year now—titled “Feeling Violation: Tactile Rendering of the Early Modern Blazon”—is an attempt at providing an alternative textual experience, one gained through tactile rather than visual interaction with a text. The driving question of the project is: can we use digital tools to measure and feel (in a literally tactile way) the affective impact of poetry? This project is really born out of what I think are very convincing arguments by early modern feminist scholars on the Petrarchan blazon and its particularly violent effect on the female body – Nancy Vickers is perhaps one of the most famous proponents of this. Here’s a quote from her essays in David Hillman and Carla Mazzio’s The Body in Parts anthology (SLIDE 2). And as we can see from images like this one (NEXT) from an English translation of Charles Sorel’s The Extravagant Shepherd—even in the early modern period, there was some thought given to this fragmenting act (although not from same feminist perspective). Following this idea, I wanted to see if we could actually measure what these feminist arguments lay out—if there was a way of rendering this linguistic violence to make it more “feel-able” to a reader of the text.

(SLIDE 3) So the early version of the project—pre-EMDA—used MALLET and Voyant to come up with a set of keywords marking violence (mostly related to body parts in the case of the blazon) and then tracked the density of those terms across the text. I split the text into roughly 50 chunks—about 325 words per chunk, so this is that “distant close reading” idea I’ve been floating around really trying to get a fine-grained analysis on the violent rhetoric (starting with one text, tracking really small segments to see how these words shift in density). I was hoping this would give me a sense of moments that were particularly violent—perhaps we could say moments with a more visceral violent texture.

Next, I used Rhino, a computer-aided design software, to convert my quantitative data into three-dimensional infographics (that’s what you see here, first plotting the data points, and then inserting poly-line segments to connect the points and give the object it’s 3D dimension). Rhino gave me a file that I could then physically render using a laser cutter or a 3D printer, extending the possibilities of data-driven comparisons and big data methods into tactile and aesthetic registers of analysis. (SLIDE 4) So here are some photos (sorry they’re not super sharp) of the objects I’ve produced thus far using laser cutting (brief explaination). And, in fact, my initial conception for what the object would look like was closer to what Stephan showed us with emoto—the sculpture installation with the textured surface mapping the depth of different topics at any given moment. But, I quickly discovered that even doing something like that on a small scale is cost-prohibitive (for a graduate student).

This project has not been without its problems and major setbacks: cost of materials being one of the big ones, but potentially more important from a literary scholar perspective—as we’re well aware, poetic language is particularly finicky, which at first I thought would be useful but I’m finding that it’s perhaps too figurative to use as a starting point for these kinds of methods (if that makes sense). A lot of the violence feminist scholars read into blazon, for instance, is in the metaphors—so it’s not just the fact that the poet mentions the lover’s eyes but that he compares them to a violent object, like a basilisk’s gaze, so it’s both the fragmenting act and these troublesome comparisons. The friction/relations between words is (as we know) very hard to map using computational analysis methods.

Shifting then into my train of thoughts over the last two weeks—my thread of questions has kept me very close to this project, but I hope that my new trajectory of ideas might lead me to slightly more successful and manageable direction for my research questions.

(SLIDE 5) This new set of ideas really started to fall into place for me at the start of Week 2, when Jonathan opened the week’s conversations asking us to think about: “what happens to texts in statistical spaces?” The day we produced our 2D and 3D models with wood and clay first got me close to own methodological penchants, but also really made me think about the weight behind each piece of data. Building also on our discussion with Lisa during one of our very first days of seminar—what is data, what kinds of rhetorical arguments come wrapped up in our data at even the tiniest units of measure, and the up close and personal look we got at “raw data” during the first week. What was each clay ball meant to stand in for and how could we think that they were in some way equal? If we count for the frequency of the word “ravish” in Titus and include this stage direction in our count, how can we distinguish, in our digital modes of expression, between “ravish” in the famous stage direction (shown here with image of Titus quarto from LUNA) and ravish in the spoken dialogue of the play. (Note: driven to this example because Titus is a chapter in my dissertation and because I’m thinking about sexual violence in texts.) How do we weigh a word in the dialogue versus a word in a stage direction that is meant to mark the body in some way, not just signaling action but signaling a past or future action that the actor’s body is then, at that moment, asked to represent. How could statistical analysis, quantitative analysis that tracks this word through Titus account for the body?

Of course, to a large extent, these are the questions behind big data and the theme of this entire two-week institute. The tension between distant reading and close reading, how we can pitch our arguments to a literary scholarly community that’s been built on close reading and analysis. But I think the ravish/ravish question was bringing me to a slightly different place – and here’s where these questions started to intersect more immediately with my blazon project, because what that question really invites me to think about is the bodies behind the texts. Like the rhetorical violence I’ve been working to map in the early modern blazon—there’s a body of the lover underneath the poetry (and, in many of the sonneteers cases we have some evidence that these were actual women they were writing to/about). But the stage direction question really hit this home for me because in performance that body is so much more obvious, so much more “there,” and that body is asked to feel and act in the ways the text directs, even when those directions are violent, or, anachronistically, traumatic.

So one of the things EMDA has helped me realize, is that I think I want to come at my inquiries using stage directions rather than poetry as a way to build a sample set and start making some initial claims about rhetorics of violence and bodily markers, and the connections to digital methods. Problem “teasing” stage practice from dialogue/text of the plays, which is primarily what EM computational analysis has provided – as fascinating as current work on early English printed play texts is, these methods seem similarly problematic in terms of gathering something about the performance and, for my purposes, something about the body in performance. (SLIDE 6) If the digital is as innovative and exciting as these past few weeks of EMDA prove, I wonder how we can use our digital tools to reveal these absences and reinvigorate the bodies of early English performance – turning from archive to repertoire in our digital practice.

As a transition, and to get us thinking about the way bodies are forced to move differently according to the stage directions, responding to these textual prompts, I want to do a short experiment: everyone please close your hands into fists – make your arms into “stumps” as the actor playing Lavinia is asked to do and try to pick up your pen, maybe even write something? I was able to scrawl “Chiron” when I practiced this during lunch, but I think it’s very useful to notice the different ways your arms have to move differently, something that ends up affecting the way your entire body controls the writing instrument as you try to get something legible down.

Here are some of the women who’ve been asked to move their body in this way—marked onstage by the ravishment of their bodies offstage (SLIDE 7) Hopkins (mid to late 18th century) and then Vivien Leigh (who was apparently not great). And then, (SLIDE 8) most famous recently I think is the Globe’s 2014 performance with Flora Spencer-Longhurst as Lavinia. As many of you well know, her performance generated quite the bodily response in her audience (SLIDE 9): fainting at the Globe performance of Titus. In large part the question my project asks, the question I want to keep at the forefront moving forward it: can we conceive of a way to even just touch this level of affective response in our digital modes? I know it’s not quite the big data scale and the questions about gradual/continuous change we’ve been talking about especially in our last few sessions, but I truly think the digital has ways of making these alternative reading and experiences of the text possible.

The first question seems to be: (SLIDE 10) where are the stage directions in digital representations of EM drama and the computational analysis that’s been done with these texts? True to Alan Farmer’s form, I started with print catalogue of Dictionary of Early Modern Stage Directions. Alan Dessen and Leslie Thomson put this together, based on electronic database Thomson compiled of “over 22,000 stage directions culled from roughly 500 plays (some of them in more than one version). While this catalogue is a great starting point, I’m not sure this dictionary is the best answer for recovering that record, those silences they refer to in their introduction. (NEXT) Print book itself includes collection of keywords followed by general sense of their use in EM stage directions, and some examples of their use; then a list by subject category, and then list of plays cited. (NEXT) Here’s the definition for “ravish” in stage directions – particularly relevant for me and Titus direction shows up as “best known.” I think digital can do a lot more with these.

Screen Shot 2015-07-06 at 2.52.25 PM

Folger Digital Texts: stage direction tag description for “business” category.

They do have an interface! (SLIDE 11) With some updates to the printed book, but the upkeep on this seems to have stopped—one of my first steps is to reach out to them to ask about their plans for this project. (SLIDE 12) So then I looked at what Folger Digital Texts are doing with stage directions. They are obviously including them in electronic editions but they’re tagging them in really interesting ways. They’ve categorized them into a few groups for tagging purposes. “Business” category (first it’s a great/weird word, second “someone will have to make them happen”) is the one I’m particularly interested in because it seems to be the one with “substantive” stage directions—what I’m calling stage directions that mark the body in some way, that give us a glimpse into/of the performing body. Obviously need to do more poking around here to see what others have done with these, particularly thinking I might move into models on more contemporary drama with heavier stage directions, seeing if we can use those models (if there are any) to look back historically.

So just as a starting point this past week—because I don’t yet have a real sense of what the end product/project will look like—I thought I would start by filtering my questions through some of the tools we’ve been working with the past two weeks to see if I could find those outliers, really hone in on what the tools are already showing us, are there any particularly interesting moments, and what might a combination of these tools do for these questions. These experiments in their initial stage, unfortunately, use the text both with and without stage directions because it’s only as of last night that Eric was able to help me figure out how to reinsert the stage directions in the plain text of the plays we helpfully got from Jonathan. Armed with this interest in stage directions and how the texts mark the performing body both through the stage directions and in textual cues around the stage directions, armed with Lisa’s excellent starting point from Day 2 (“what’s missing from our data,” “our digital systems are very bad at marking what’s been left out”), and with Jonathan’s provocative presentation at the start of Week 2—especially the question of data reduction in digital space—how does that flattening of dimensions of a text in something like PCA/big data computational analysis further reduce the body behind the numbers?

Starting with AntConc (SLIDE 13), just to look at some of these words in context, I was reminded that, not only do we have some crucial stage directions in Titus that mark the actor’s bodies in important ways, we also have Rape figured onstage in this play. And this ends up being the perfect example of how these tools help us see important elements with a more finely grained analysis, but also that these tools can/will flatten out the complexity of those details if we let them. If we count frequency of the word “rape” in Titus and start playing with different modes of analysis to visualize a rhetoric of sexual violence, for instance, that happens across the play, we miss something like this: such a crucial moment where the word “rape” becomes the name of a character. How can we show the difference using quantitative methods between rape as uttered word and rape as figure? Seems to me that quantitative methods, visualizations in this case do indeed flatten out the complexity behind this shift—getting back to our conversation on Monday afternoon with Ted and Andrew.

Then I moved onto Docuscope (SLIDE 14)—and this is running the texts without the stage directions, but still a useful experiment to see what Docuscope gives us for moments directly adjacent to substantive stage directions, where the body has been asked to perform in a heightened manner. This is 2.4, just after the “ravished” stage direction I showed you several slides back. So the dictionary built into Docuscope tags as “motions” a lot of the action words, but these often seem to be metaphorical or conceptual actions rather than actions onstage (and, indeed, Jonathan confirmed this when I asked him about it yesterday). And then it tags as “negative emotion” some of the phrases that actually mark the actor’s body—here we see “lost her tongue” is tagged that way. That’s something to think more on: why the structure of the strings here (because it is catching this whole phrase together), why the dictionary reads it functioning as an emotion instead of a bodily marker. And perhaps most sadly, it doesn’t tag the instances of “ravish’d” here in the dialogue at all (although I think that might be because of the final ‘d – I need to run it again with modernized endings perhaps).

Docuscope: dialogue at 2.4 tagged according to Docuscope's dictionary and perceived rhetorical functions.

Docuscope: dialogue at 2.4 tagged according to Docuscope’s dictionary and perceived rhetorical functions.

Finally landed back at topic modeling, which is where I started my blazon project, and I think the implications of that return have some significance for the promise of this tool for me. (SLIDE 15) So this slide is showing three of the most salient topics generated for Titus (and this is using MALLET, but through Eric’s great Serendip interface for helpful visualization). The stage directions are in place here, so the topics are being generated with those words in the mix. As we can see here, the most prominent topics for Titus do indeed register a rhetoric of violence that spans the entire play (we don’t need topic modeling to tell us this). And, perhaps as expected, the stage directions are so sparse that they make little impact on the topics (so, this is again something to think about—topic modeling as a method and the “weight” of different textual markers).

But what is crucial for me here: running the text with or without stage directions, the violence to Lavinia’s body is completely elided by the more “masculine” registers of violence (both in the text and in the topic modeling process). The most salient topics like these don’t immediately mark sexual violence (with the terms blood/hands we get a tiny glimpse of Lavinia’s trauma—but in the text, the hands explicitly referenced tend to be Titus’, when Lavinia’s are mentioned they’re referred to as “branches” and “limbs”). And “ravish” and “rape” don’t make it into these topics at all. Perhaps in part this usefully ties Lavinia’s violence to Titus/family, but sexual violence, which is arguably at the play’s very core (the element that makes the Globe audience faint), is ABSENT!

Serendip: rank viewer for saliency of word

Serendip: rank viewer for saliency of word “ravish” throughout Shakespeare’s corpus.

(SLIDE 16) We do see traces of “ravish” in the rank viewer in Serendip—so the red lines here are marking the saliency of a word in a given topic. The topics that “ravish” is salient in are generally the same ones that are most dense in Titus as a whole. So this is showing you a close up of Topic 3, where ravish is the most salient than in any other topic, BUT ravish doesn’t make it into the top word list on the right here (it’s “chopped” by the program’s algorithm). And, instead, we get these terms that mark, again, more masculine forms/modes of violence: blood, rage, bloody, enemies, deadly.

We see this even more convincingly when we look at “rape” (SLIDE 17). It only shows up in one topic across Shakespeare’s entire corpus, and this is again another topic particularly prominent in Titus. And the red bar this time is very high up in the topic—so this word is very salient in Topic 5, much more than ravish is in any given topic—but again it doesn’t show up on the topic word list to the right. Where does it go? So even when these words are heavily marked by the texts, they’re still covered over. And, again, what this is not catching is Lavinia’s body as a signifier of sexual violence throughout the play, or the way “rape” eventually becomes a figure/character onstage rather than simply a word in the dialogue.

Serendip: topics across span of Titus with colored lined indicating topics marking more violent language.

Serendip: topics across span of Titus with colored lined indicating topics marking more violent language.

As a final example (SLIDE 18), here’s a close-up on the topics most salient across the span of Titus. I’ve marked the ones I think are tied especially to “violence” (rhetoric of violence). I think the blue line is particularly interesting—this topic again seems to signal ways the performing body is marked by violence, but it perhaps has more to do with reactions to the violence performed on others’ bodies: weep, tears, sorrow, cheeks, woes. So if we track this topic, it spikes very high at two key moments in the play. The first high blue spike: Titus’ speech on boundless tears (weeping as performing action), as his sons are carried off with the senators to be executed (this is right after 2.4 but Titus hasn’t seen Lavinia yet, Marcus is about to bring her to him). And then the second high blue spike: at 3.1, just after Titus has chopped off his hand and he’s then brought the heads of his two sons. Which means that his reaction to Lavinia’s marked body after her ravishment is tucked between those two spikes—still high, but in the valley (and this is not a tool problem perhaps but an instance where I’d want to go back to the text and look at these moments to see how Lavinia’s body and Titus’ reaction to it are coded in this moment).

As I wrap up, just a few brief thoughts on where I want to go moving forward from EMDA. First, on a critical/theoretical level, I want to go back to the feminist frameworks I started with for my blazon project and think more about how I can leverage feminist methodologies for this project. I spoke to Ted Underwood on Monday briefly about what kinds of evidence we can pull from feminist critical modes, and how they might provide us with a way out of the evidence dichotomy we’ve been talking about: big data vs. anecdote. I also want to turn to performance studies and how the field has dealt with stage directions, recapturing the performing body and Diana Taylor’s repertoire. Digital methods could greatly benefit from the way performance studies has dealt with silence, gaps, absence, etc.

On the digital side of things, I want to think more about topic modeling—this unsupervised analysis method that, as we’ve posited, provides a glimpse into the “latent content” of the texts we study. If that’s the case, then I want to use the examples I’ve laid out today from Titus as a starting point for a rigorous examination of the algorithms behind this method. Our arguments from close reading would suggest that the “latent” content of Titus lies in a very different place than these topics demonstrate. What do we do with that tension? I also want to explore Mitchell Whitelaw’s idea of the generous interface—if this project will eventually result in a digital archive of stage directions that tries to demonstrate some of the problems and instabilities that I’ve laid out here, is there a way for the interface to match that? Could I build an interface for this project that embraces the ephemerality, the silences and absences marked by the stage direction?

And, finally, building these questions and this new focus back into my original project on feeling violation, I want to think about ways to get our bodies moving and feeling in relation to the performing bodies of these texts. This could be on the level of 3D objects that ask us to read and feel the text differently, but it could also be experiments with digital/physical renderings of audience reaction, bodily simulation—there are a lot of avenues for these bigger questions. And with that I’ll end and just say I’m excited to move forward with this and present some exciting new work during our reunion weekend.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: