Gephi Put Me in My Place

My experience with Gephi, while not one full of cursing and outward frustration, was not entirely pleasant. I say it was not entirely pleasant because all-in-all, Gephi is a unique tool that--like Voyant Tools--opens new doors for me as we trudge through the semester and make our way to the deadline for our semester-long research project. I am happy to have been introduced to this tool, but being unfamiliar with the data loaded into it was most definitely a source of frustration for me. Had I compiled the data myself to load into this, that may have made the blind exploration process easier as I know how the data is supposed to look in certain situations.

The genre tags made little sense to me without context of how they were developed, but I'm sure made sense to whomever compiled them. I enjoyed playing around with the different types of graphs and applying different styles to the font, which makes me excited to try Gephi again with my own data. Something that was also interesting was reconfiguring the data to highlight different trends in the graph.

I loved exploring the statistical tools on Gephi, regardless of having no idea what the results implied with little knowledge of the original data. Gephi was definitely foreign to me, which is strange because I typically find myself as the pseudo-IT person in most situations as my professional work includes a focus in web devleopment. It made me uncomfortable having little control over what I was doing, but I enjoyed that discomfort and figuring out how to manipulate the graphs. I look forward to employing this tool again in the future with data that interests me.


"I think I'm getting the hang of this Gephi thing!"—My famous last words. 

Anyways, the potential of Gephi and data mapping in regards to literature is, honestly a little thrilling? It's a whole different way of reading works that we've read for ages. Like I mentioned in class, the example work Gephi uses to demo their software is characters in Les Miserables. It might be too much to say that the potential of software of Gephi it makes these works feel new, but the idea that there are overlying structures that we don't usually see and now can... That's amazing! It's like we have X-ray vision. 

Now, applying the software to my data proved more challenging. I think my nodes might have been broken somehow, since I only had twelve, and it was easy to make mistakes that I didn't know how to undo, but following the step by steps from Gephi and Dr. Drouin, I think I was able to at least sort of arrange the data how I wanted. It had shapes and colors, and when you export as a PDF, it does this beautiful curve thing. Quite nice to look at. 

The thing I guess I am still so intrigued/frustrated by, is how it feels difficult to know which questions are the right to ask. I guess in the humanities I'm so used to the more philosophical and theoretical side of things it feels weird to look at something this mathematic. Once you have a graph organized and layed out is when the questions really start to arise. In this case, it really layed out the thematic soup all the authors were living in. Relatable! 

I tried to embed my photo, but I don't know that it succeded so I am also linking a photo to my work. 

Every Issue Is a Themed Issue

As an editor of a literary journal, you can't go much lower than a themed issue. It's the absolute zero of the creative writing world. An editor's sad nadir. A literary journal's last death cough from a consciousness that's blinking out. The other day, an email arrived in my inbox from some literary journal straddling uneasily the periphery of relevance. They were advertising a new call of submissions for a scab themed issue. The advertisement read:

Not quite broken, not quite mended. A reminder of what was, and what will be. Scabs protect our wounds, and yet the temptation to pick at them and peel them off is always there. They call attention to what is already disappearing, and we don’t yet know if they will leave a scar. Scabs are a reminder that healing can be a long, ugly process.

It's the kind of thing that should make any creative with dignity left blush. Yet, it's also kind of instructive, right? It announces what we already know: a lot of people are in pain right now, so a lot of people are writing about pain. Scabs are wounds. Scabs are painful. Scabs are a kind of healing. Scabs are ugly. Pain-->scabs-->wounds-->healing-->ugly.

What I’m trying to say is that Gephi allows us to access these kinds of associative networks that we are always consuming as we read. And we know exactly what we are consuming, but we don’t talk about it, because it’s obvious and kind of embarrassing to talk about. Ugly correlates with scab, because scabs are ugly.

The fact I spent my time trying to operate Gephi with all of the subtlety of a baby trying to fit the triangle shape into the square hall isn’t the point. The point is clear. Identifying associative networks is requisite element of performing close reading is. Gephi just allows us to access what we already know but can’t express yet. Turns out: every issue is a themed issue.

Gephi analysis

I think there was some kind of error in my data importing process. While I agree with the topic tag connections, there are duplicate tags that will list different permutations of the same words. I would have to do more investigation to figure out how to eliminate the duplicates. They make the graph more cluttered than I would like.

The Yifan Hu layout algorithm seems to bifurcate the data based on the amount of connections between tags; in the chart that I’ve created that filters the ID network by relationship to James Joyce, the network produced by the Yifan Hu layout is two wings connected by a central node of the novel, which is a convenient visual to conceptualize the two main written works related to Joyce. When the full data set is arranged in the Yifan Hu layout, the duplicate data creates a lot of visual clutter as the data separates into grouped islands; I’ve included a screenshot of one island, where I can note the centrality of the term “poetry,” although the rest of the data is a bit jumbled.

Given the limits of visuality in the Yifan Hu arrangement, I was pleased with the image generated by the Fruchterman Rheingold layout, which created a modular arrangement that flexes upon rearrangement to maintain the relationship between nodes. I felt that this model was more visually intuitive (although that may be partially because of the visual limitations of the duplicate nodes) because it kept the data contained and connected rather than dispersing it into discrete islands, which I think models the conceptual relationships between topics.

I think this network visualization is a useful method for periodical studies (given the future ability to import data without error…) because it reveals relationships that are not evident upon close reading, but, once revealed, may help guide/inform close reading. I think I could benefit from this procedure as a way of giving a fresh set of eyes to the same brain.

Never Have I Ever...

...felt so technologically inept as I do now while trying to navigate all these fun new tools (*cough, Gephi) I never knew existed, so I might as well blast all my shame here for everyone to see. Once upon a time I might have rated myself as technologically proficient (although even that seems like a stretch now), but I have permanently disabused myself of any notions of ever being 'tech-savvy.' I used to laugh when my Dad struggled with basic computer usage because I didn't get how hard it was for him to navigate unfamiliar territory, but I'm sure not laughing now (sorry Dad!). 



To add insult to injury, that blank space ↑ is where the picture of my Gephi Little Review is supposed to appear; apparently I don't even know how to share an image of how much I don't know how to use Gephi. (I uploaded it to Shutterfly & copied the link because I don't know how else to get an image url.)  

I have muddled my way through Gephi as much as possible, but I honestly have no earthly idea what I'm doing and nothing really means much. I can't tell if there's anything missing, or how the nodes really relate to each other through the edges; because of this, I can't offer an educated opinion on the genre and topic connections. I thought the Fruchterman Reingold layout was more aesthetically pleasing than the Yifan Hu layout, but I didn't really glean anything different from one or the other. Despite this, though, I think Gephi is a veritable treasure trove of potential for literary analysis of all kinds; the samples show how well it works with Les Mis and it would be really fun to see Moretti run his Hamlet analysis through Gephi. I get the idea even if I don't actually get it. 

(And now, if you'll excuse me, I'm going to hide in a corner and try to figure out how I got technologically left behind in only my thirties...)

Gephi part II

(Note: I will e-mail pics to professor as they will not download here.)

3) The Fruchterman Rheingold program changes the scheme into an almost hexagon like shape. In this simulation, the inner words are prose and irony, and they are surrounded by all the other words that intersect in this formation in the photograph. It leads to another kind of meaning, because of the nuance when these words are mentioned, also bring eventually these other words into the limelight. This is another layer to explore and quite helpful.  

4) I can see the advantages of working small and medium sized visualizations. Smaller ones would be useful in the case of zeroing in on certain aspects of research, narrowing the focus of a single subject or word, for instance. Medium sized visualizations would work better for larger groups of words or showing how graphs can change over measurements of time. Measuring the data of literature is not a business set in stone, though. I suggest incorporating these tools, but to not wholly rely on them to get all the answers. Updating models will be paramount as technology keeps changing and better tools are developed. Even more importantly, close readings still need to be maintained for what machines miss that only humans can see.  




Gephi answers and pics p

1) Yes, I did notice a difference when playing with the system. Based on the magnification of focus, words would disappear from the word structure. I disagree with the structure based on level of importance of words. ‘World War I’ is the most important and should be in the center because it effects all the words around it more than any other. Even when the word may not be said, or said as much, it is implicated in other terms and phrases because it affects everything. When other words are used like death, memorial, greatness, etc. In the picture. However, all these topics point back to WW I as the instigator. This cannot be gleaned from the graphs, but only close readings of the material. 


2) Based on the running program for Yifan Hu, the shape of the diagram changed drastically. The shape changed into a kind of rhombus that kind of tilted from slightly northeast to southwest, with Memory of Gregory at the top. Mediocracy, world, and Death meet each other more often in the center of the shape, showing how often they intersect with one another. I find this image more in depth and able to read the author’s true intent by re-aligning words in a different structure, making them more ‘visible’, or in other words, more obvious. 


White and Children

Whenever I teach English Composition, I tell students that good writing is really only a matter of building bridges only you can build; making connections no one else can make is what separates an essay that is a rote mechanical exercise of getting words on paper from a creative expression. One day I will write a paper about Joycean neologism in Finnegans Wake and neologism in the music of Young Thug. But that’s for another response. Point is: when I went into Voyant for the first time, I was excited all of the new avenues of bridge building that the digital humanities provide. I was drawn immediately to the contexts and correlations sections at the bottom of the window. The correlation that interested me the most was between the words “children” and white.”

It’s a curious correlation for a number of reasons. In a journal made by and for black Americans, why are white children appearing so often and so close to one another? I can’t help remembering my disgusting priest’s classroom in my wasted Catholic high school. A pro-life poster that read “black children are an endangered species,” was the only permanent fixture on his only bulletin board. We passed it every time we were dismissed. Dear God, a species. Of course, no analogous poster exists for white children. A bias, even an unconscious one, is obvious: white children are, well, children. Black children are not. Consider Tamir Rice, the twelve-year-old boy who was shot for carrying a toy gun. For black children in this country, childhood and play ends prematurely. As soon as black bodies appear within the institutional gaze, childhood gives way to a long adulthood. Now, obviously I’m reaching, but I think that’s the while point. The process of the reading and building the digital humanities not only allows for, but encourages reaching and grasping for ideas that are just beyond the periphery of traditional paper scholarship.

Reading The Crisis with Voyant

Looking at The Crisis through Voyant was a novel way of reading through text that ultimately proved to be quite difficult. In particular, I found it hard  to know the right questions to ask. 

The thing about the sort of close readings I am used to, is that your brain is the machine, indexing and interpreting in ways that we don't think twice about. It’s like walking versus driving. You can do things with a car that you could never do by just walking, but you have to adapt your thinking to the vehicle's mode of transportation. The kind of connections that we make through reading the text have to be manually input by us, yet, without satisfactory knowledge of the material I was feeding Voyant, any question or hypothesis felt like a stab in the dark. 

I wanted to know if the language surrounding race changed over time in The Crisis, and so I plugged in the words "negro" "colored" and "black", and my results were... inconclusive. I had assumed that as time went on, negro and colored would go out of fashion and Black would become far more popular, but that wasn't the case. Even futher, there are no results for "African American", I hadn't realized how recent the term was.

Overall, negro and colored were both used seemingly interchangably, with black being far less used overall. in the last few data points, however, use of colored and negro is down, while black remains the same. I'm not sure what exactly to make of any of it. 

Regardless of how much more removed I felt, it turns out, distanced textual analysis still requires one to get their hands in the mud and grapple. Reading Crisis through Voyant only convinced me of the need of both distanced and closed reading working together in tandem for more complete picture. 

Distant Reading of The Crisis with Voyant Tools

Considering the range of publication years of The Crisis, I used Voyant Tools to see the woman’s suffrage movement fused in the magazine along with the race issues. Since it was not a one-time movement, but a long-term one, I found it super helpful to track the movement within the circle of The Crisis columnists and audience, using Voyant Tools which extracts the results across the whole series of magazine issues. I first typed in “suffrag*” and saw the five peaks—04.5 being the highest, then 10.4, 15.1, 11.1, and 12.3 in sequence—in the graph that deal with the suffrage movement the most. Thinking that it is not as often as I imagined, I became curious about how frequently The Crisis deals with any issues pertinent to woman and typed “woman*.” As it’s shown in the graph below, the top three peaks coincide with the results of “suffrage*,” which reveals that one of the main social issues The Crisis printed in relation to woman is the woman’s suffrage movement, at least before 1920s.


Meanwhile, I became curious about whether the suffrage issue dies down after the 19th Amendment, and so I typed in the words, “disfranchise*,” “enfranchise*.” The result below shows that “disfranchise*” has a highest peak in 21.4, which seems to speak that after the enfranchisement for all people, the magazine is harshly critiquing the issue of disfranchisement still existing out there.


The experience of distant reading with Voyant Tools definitely helped me have a fresh perspective to see the construction of the context around the readings in longer period.