CC BY NC ND-licensed photo “stones 2″ by Flickr user speedoflight_speedoflight. Source

It is, of course, absurd to claim you can capture the richness of human experience in machine-readable data. Human lives are quicksilver, protean, bent and pulled in a thousand different directions. We think and feel, interpret and surmise, hold contradictory notions, revel in paradox. It’s ridiculous to think that a machine, which thinks in binary, can replicate these shades of gray.

And yet. Media scholars know better than anyone that it is equally absurd to attempt to capture human experience in a photographic narrative. Because we understand the photographic image — its trickery, its inherent limitations, the world beyond its frame — we understand how essentially false is any work’s claim to represent “reality” in all its plenitude and contingency. To argue that a work of media is fully representative is to be unforgivably naïve; we know that every work is constructed, no matter how transparent it appears.

But somehow we feel there’s something valiant in the attempt to capture human experience, even in these inadequate media. Writing on Rossellini’s Paisà, Andre Bazin observed that the film’s essential unit is not the shot but the “fact,” one slice of time and space, itself worthy of interpretation and filled with meaning. Paisà is rife with gaps and omissions, but so much the better: “The mind has to leap from one event to the other as one leaps from stone to stone in crossing the river.” The best films are beautiful not because they claim earnestly to represent reality, but because they acknowledge this feat’s impossibility butkeep trying anyway, honoring their viewers by trusting them to make their way from stone to stone.1

There’s a potential for a digital humanities that holds toward data the same vexed, impossible loyalty with which media scholars honor the photographic image. In this version of digital humanities, scholars would view data neither as fully adequate to reality nor as necessarily mendacious, but as one moment, a slice of time and space. The best work would not be the most comprehensive — just as the best films are not the most verismilitudinous — but that which exhibits the most sophistication, the most humanity, in making the leap from fact to narrative.

I don’t think digital humanities is there yet, but I think this is an opportunity for media scholars. This is why I think the best possibilities for the intersection of digital humanities and media studies lie not so much in counting frames or automating facial recognition (though this is interesting in its way) as by bringing to digital humanities the peculiar agony of the media scholar: the belief, simultaneously, that all stories are lies and that there’s truth in their telling.

1 André Bazin, “An Aesthetic of Reality: Cinematic Realism and the Italian School of Liberation,” in What Is Cinema? Volume 2, André Bazin (Berkeley/Los Angeles: University of California Press, 16-40), 35 and 37.

My office at UCLA’s Center for Digital Humanities

If you’re curious about what I do all day (and I actually do get that question a lot), I’ve documented my day here, as part of the Day of Digital Humanities project.

A decent daily-use electric breastpump like the Medela Pump in Style will run you $269.99 on Amazon.

Breastfeeding has been a pretty damned-if-you-do, damned-if-you-don’t experience for me. I’m in an extremely privileged position, breastfeeding-wise — with relatively generous (for the U.S.) maternity leave and a private office with a door — but it’s still been a challenge. New mothers hear a great deal these days about the expense and health toll (though frankly some of that science is questionable) of formula-feeding (or, as Kaiser’s lactation consultant insisted on calling it, “artificial food”). But breastfeeding also has well-documented and significant financial penalties for women who work outside the home. And people who wouldn’t ordinarily pronounce on a woman’s personal decisions feel no compunction, for some reason, about passing judgment on a mother’s decision about how to feed her baby.¹

At the moment, I’m on my way back from the Society for Cinema and Media Studies, my first real trip away from the baby. The conference organizers were really helpful to me when I asked for lactation accommodation, finding me a room in the sold-out conference hotel so that I wouldn’t have to miss too much of the event. (Though this, of course, meant staying in the conference hotel, which I usually avoid in order to save money.) Still, it’s been a bit of a logistical challenge, involving trekking across airports in search of remote nursing rooms, sending a pump through security checkpoints, and absenting myself from events every few hours.

In a perfect world, we’d all understand the mechanics of lactation so that we can accommodate women who are breastfeeding. But I’d be pretty wildly hypocritical if I condemned others for their ignorance, having until recently been in the same position myself. A few months ago, I was embarrassed to realize I had no idea where to send a woman who needed to use a breastpump at an event I’d helped organize. I didn’t even really know what she’d need, having never dealt with it myself. Which is to say that I understand why this stuff is confusing. So I thought I’d do my tiny part by explaining why we need what we need for the benefit of anyone who might host a breastfeeding mother.

Maintaining a supply of expressed breastmilk is a complicated operation, involving considerations of supply, demand, inventory, expiration dates, stock rotation, storage space, and transportation. I use an iPhone app called, mortifyingly, Milk Maid, to help me keep it all straight.

Most women who pump frequently use electric breast pumps. So a breastfeeding mother will need, at a minimum, a private room with an electrical outlet. Public restrooms won’t work because they’re not very private, they may not have an electrical outlet, and they’re not very sanitary. Plus, frankly, it’s unpleasant to hang out in a public restroom for 20 minutes at a time.

It’s really nice to have access to a fridge (even nicer if it has a freezer) to store milk, although many women have coolers for breast milk. It’s also nice to have access to a sink to wash bottles and pump parts (and dump milk if necessary).

It’s also very helpful if I can pump at the last possible moment before I travel, because flights can be long and many airports don’t provide lactation rooms. This may mean extending my hotel checkout time, an accommodation that unfortunately proved impossible at the SCMS conference hotel. If the conference organizers had said something to the Drake Hotel’s management, it would have been easier to make my case at the hotel’s reception desk.

So that’s all. It’s not too complicated. It’s really nice when someone anticipates these needs — as Jacqueline Wernimont did for this weekend’s THATCamp Feminism — so if you’re hosting a large group or you know that an attendee has a young child, you might just include a line about lactation accommodations in event materials or an email, the same way you offer accommodation to people with disabilities. As a matter of principle, I make a pretty vocal fuss about these things, but you can imagine why raising this issue is uncomfortable for many mothers.

And, not that you would do this, but it has happened to me: Unless you know me really well, please don’t make a hilarious joke about breastpumping, recoil in theatrical horror when the subject comes up, or chide me for grossing you out. It makes me feel weird and gross, and heaven knows I’m not doing this for kicks.

^{1. The American Academy of Pediatricians recently advised physicians that “infant feeding should not be considered as a lifestyle choice, but rather as a basic health issue,” a statement that I find breathtakingly hostile to the realities of breastfeeding for many women, given its significant emotional and financial costs.↩}

Visualization of all close-ups of faces from Vertov’s The Eleventh Year by Lev Manovich, in “Visualizing Vertov.”

This is the introduction I gave to a workshop on media studies and digital humanities at the Society for Cinema and Media Studies annual conference in Chicago on March 8, 2013. Fellow participants: Eric Faden, Hannah Goodwin, Jason Mittell, Jason Rhody, and Jasmijn Van Gorp. Many thanks to the SCMS livestreaming team. You can view the taped event here.

In 2000, the media scholar and digital humanities practitioner Johanna Drucker sat on a panel at SUNY Albany with Jacques Derrida. They were there to discuss digital media, but something totally unexpected happened: failure. Derrida was “unable,” writes Drucker, “to get a purchase on digital media.”¹

The problem was not Derrida, but theory itself. Derrida made his observations at a remove, pronouncing at a distance on the changes wrought by digital technology. “This will not do,” Drucker declares, not even for one of the greatest theorists of our time. We must theorize digital technology through critical engagement with the medium itself, through making and breaking and building and reflecting. Pressing the humanistic against the digital, acknowledges Drucker, we fail and fail and fail, and “what is revealed in the processes is not what the machine does not know — but what we have not, until this exercise, been ourselves able to see.”²

Film “fingerprints” created by Frederic Brodbeck using his Cinemetrics software.

Thirteen years later, media studies remains largely absent from the array of disciplines that have staked a claim on digital humanities. I say this with two important caveats: first, that, as witnessed by today’s workshop participants, some media scholars are doing digital humanities; and, second, work that we might be happy to claim as digital humanities goes on under different names. Alex Juhasz, for example, does work that might be called digital humanities, but has chosen instead to ally herself with feminist traditions of activist media.

But the fact remains that SCMS has yet to host a panel or workshop explicitly devoted to digital humanities — evidence, one might venture, of a lingering divide in the discipline between theory and praxis. Or perhaps media scholars, so expert at interpreting the ideological effects of media objects, understand too well the dangers of engaging flippantly with digital media. Perhaps we’ve just needed time to think about it.

Jason Mittell and I decided to force the issue by putting together this workshop, with the aim of seeing what happens when media studies and digital humanities converge. Our aim is not to argue for any particular approach to DH and media studies, but to gesture toward what might be done at the intersection of the fields. To that end, we’ve assembled some people who are doing interesting research, and we’ve asked them to keep their presentations brutally short, so that we can reserve most of the time here for questions and discussion — in which I hope you’ll see yourselves as much participants as those of us at the front of the room.

Shot lengths for Charlie Chaplin’s His New Job, created using Yuri Tsivian’s Cinemetrics database. Read Matt Hauske’s analysis here.

How one defines digital humanities is so contentious that anyone who does DH has to suppress a groan when the question arises, as it inevitably does. But since I’m the one doing the introducing, I’ll share with you my definition, which is a simple one: it’s the use of digital technologies to investigate humanities questions. I say “investigate” and not “answer” because it is eminently clear that one cannot definitively answer a humanities question. Definitive answers make no sense for the humanities, which deals in and excels at ambiguity, paradox, and shades of gray — all qualities at which, I think we can agree, digital technology fails miserably. So why apply the one to the other? Because it’s the friction between the two that’s productive. From moment to moment, as we answer email and check Facebook, we already swim apparently effortlessly between the realm of the human and the ruthlessly binary logic of the digital. Digital humanities seeks not to bridge a divide but to trouble an already collapsed distinction, bringing the two ways of knowing into a productive dissonance.

And such a paradox should not, after all, seem so unfamiliar to film scholars. We know better than anyone that reality consists of a plenitude, contingency, and variety to which film can never be adequate. The messy and nonsensical world rushes past us like a river, in Bazin’s familiar formulation. We know our efforts to hold it in our hands are futile. And still film tries, and still we remain faithful to it, because, paradoxically, it is in the bending of reality to artifice that we witness some of its most arresting truths.

Screengrab from Mark Hansen’s shi jian: time, published in Vectors Journal of Culture and Technology in a Dynamic Vernacular 3:2 (Summer 2012).

It’s tempting to see some looming significance in the counterprogramming of this workshop with the close analysis workshop across the hall. But I caution us against such easy dichotomies, in which we figure digital humanities as a new episteme to replace the old. Instead, I urge us to see this as an opportunity to draw on those qualities at which media studies excels — the ontology of the image, a nuanced understanding of indexicality, an aliveness to the variegations and ambiguities of spectatorship, to name a few — and to ask what they can bring to the digital humanities. We will, of course, fail to find answers, but that is, maddeningly and inevitably, the point.

^{1. Johanna Drucker, “Theory as Praxis: The Poetics of Electronic Textuality,” Modernism/Modernity 9:4 (2002), 683–691, 683.↩}

^{2. Ibid., 688.↩}

Dora. Photo by Brad Wallace.

Andy and I welcomed our first child, Dora Joan, on December 15, 2012. She’s absolutely amazing. As you might expect, my world has been very much taken up with baby stuff lately. I’ll be on maternity leave until UCLA’s next quarter begins on April 1. I hope you’ll understand if I’m not great about getting back to correspondents right away.

Dora’s namesakes are her great grandmothers, whom we loved very much: Dora Goldstein and Joan Hildebrant.

Creative Commons-licensed photo by Niklas Wikström.

These are notes and links for a talk I’m giving on digital humanities and sharing your work at the University of California, San Diego, on November 5, 2012.

DH projects I discuss

• The Real Face of White Australia, by Kate Bagnall and Tim Sherratt. For more on this project, see Tim Sherratt, “It’s All about the Stuff: Collections, Interfaces, Power, and People.”
• “A Report Has Come Here”: Social Network Analysis in the Papers of Thomas Jefferson, by Lauren Klein.
• Mukurtu, directed by Kimberly Christen.
• Mining the Dispatch, by Rob Nelson. (For more on topic modeling, see this post and the related resources.)
• Complex Television, by Jason Mittell.
• The Nicest Kids in Town, by Matthew Delmont, and the Scalar platform.

Introductions to and resources on scholarly communication

• Open Peer Review: A Study of Contexts and Practices
• New-Model Scholarly Communication: Road Map for Change (PDF)
• Fight Club Soap, by Bethany Nowviskie
• SHERPA ROMEO (for learning about different publishers’ and journals’ copyright policies)
• Using the SPARC Author Addendum to Secure Your Rights as the Author of a Journal Article

Some guidelines for your online persona

• Be generous!
• You needn’t sanitize yourself, but think carefully about your boundaries.
• Never speak ill of an employer.
• Be consistent.
• Learn the conventions of online communities.

Online communities to explore

• Academia.edu
• LinkedIn
• Twitter
• Tumblr

Ways to share your work

• WordPress (my instructions for getting started)
• Omeka
• Zotero
• SlideShare
• Internet Archive

Making your work compelling

• Images
• Avoid jargon
• Edit!
• Cross-post
• Consider a less-restrictive license
• Engage with others’ work

Resources on developing a professional online academic persona

• Creating and Maintaining a Professional Presence Online, by Ryan Cordell
• Do You Need Your Own Website While on the Job Market?, by Jentery Sayers
• Creating Your Web Presence: A Primer for Academics, by Miriam Posner, Brian Croxall, and Stewart Varner

On where to start in DH, see Lisa Spiro, Getting Started in the Digital Humanities

As Zoe Borovsky brilliantly demonstrated when she visited my DH grad class, topic modeling starts with the assumption that each document is made up of multiple topics — like lumps of Play-Doh. Photo: “Play-Doh” by dbrekke.

If you’re reading this, you may know that topic modeling is a method for finding and tracing clusters of words (called “topics” in shorthand) in large bodies of texts. Topic modeling has achieved some popularity with digital humanities scholars, partly because it offers some meaningful improvements to simple word-frequency counts, and partly because of the arrival of some relatively easy-to-use tools for topic modeling.

MALLET, a package of Java code, is one of those tools. It’s not hard to run, but you do need to use the command line. For those who aren’t quite ready for that, there’s the Topic Modeling Tool, which implements MALLET in a graphical user interface (GUI), meaning you can plug files in and receive output without entering a line of code.

David Newman and Arun Balagopalan, who developed the TMT, have done us all a great service. But they may also have created a monster. The barrier for running the TMT is so low that it’s entirely possible to run a topic modeling test and produce results without having much idea what you’re doing or what the results mean.

So is it still worth doing? I think so. Playing with the results by altering variables and rerunning the test can be a useful way to get your head around what topic modeling is and isn’t. And, as I recently tried to convince my graduate DH class, screwing around with texts — even if you’re not totally sure what you’re doing — can be a surprisingly effective way of getting a new perspective on a body of work. Finally, seeing how many decisions need to be made about  texts and variables is a great way to understand that topic modeling is not a way of revealing any objective “truth” about a text; instead, it’s a way of deriving a certain kind of meaning — which still needs to be interpreted and interrogated.

But in order to get any of these benefits from the Topic Modeling Tool, you need to be able to make some sense of your results, which is no easy task. The TMT generates some decidedly cryptic-looking files, and as far as I can tell, there aren’t many resources out there to help you make sense of them.

Once you survey the results of the Topic Modeling Tool, it becomes clear why topic modeling often goes hand-in-hand with visualization. The format of the results makes it difficult for a human being to discern patterns in them, and the files aren’t easy to visualize without doing some custom coding.

But say you’re a non-coder using the Topic Modeling Tool to screw around. You feed it some text, you get some files; now what?

What follows are some very basic ways you might begin looking at the results you’ve generated.

For the purposes of demonstration, I’ve used a set of 3,584 emails that constitute the years 2008–2012 of the Humanist listserv. We downloaded the emails here and then divided each volume into individual emails. You can find the dataset I used here, and the files that comprise my TMT results here.

For further reading (and viewing) on topic modeling, I’ve listed my favorite resources here. For more on the Topic Modeling Tool in particular, I recommend this summary and video of a talk by David Newman, along with the accompanying slides (PDF).

What are these files?

When you feed a body of text to the TMT, you get two folders: output_csv and output_html.

CSV stands for comma-separated values. The documents inside your output_csv folder are spreadsheet documents that usually open in something like Excel. HTML documents will open in a web browser.

You have three CSV spreadsheets:

• DocsinTopics.csv, which provides you a list of topics and shows you which documents they’re likely to appear in;
• Topics_Words.csv, which offers you a numbered list of “topics”; and
• TopicsinDocs.csv, which provides a list of documents, along with the topics that appear most prominently in each.

You have one main HTML document, called all_topics.html. As we’ll see, this offers a numbered list of topics, along with a way of drilling down into each topic’s associated documents. Also inside your output_html folder are a folder called Docs, which contains an HTML page for each document; a folder called Topics, which provides an HTML page for each topic you’ve generated; and a document called malletgui.css, which provides your web browser with some instructions for displaying each HTML page.

Start with TopicsinDocs

I find it useful to start with the TopicinDocs.csv file. But the file requires a bit of explanation. For one thing, it’s helpful to know that each row is meant to be read across.

• In the image here, Column A gives each document a number.
• Column B provides that document’s filename. (We’ll make that easier to read in the next step.)
• Column C tells you which topic (from a separate numbered list, which we’ll find in a different document) is represented most prominently in that particular document.
• Column D tells you what contribution that topic makes to the document in question. For example, for document 1, we can see that Topic 9 makes a contribution of 0.384 —or 38.4% — to the document’s contents.
• Column E provides the next-most prominent topic in the document (in this case number 27).
• Column F tells you what contribution that topic makes to the document’s contents.
• Column G provides the next-most prominent topic …

… and so on.

Make filenames easier to read

Your spreadsheet is trying to be helpful by providing you with the entire path to each document in Column B; that is, by showing you how to navigate to each document. But that’s hard to read. I find it helpful to get rid of most of the path information — in this case /Users/miriamposner/Dropbox/UCLA/DH 201/Topic Modeling/Datasets/Humanist emails 2008-20012/By individual email/ — by selecting Column B and using Excel’s Find and Replace function (in the Edit menu) to get rid of every occurrence of that path.

Once you do this, you should be left with just the filename of each document, which is much easier to read.

Get a sense of what each row means by making a pie chart.

The kind of topic modeling that we’re doing assumes that every document contains multiple topics. That’s why each row lists multiple topics for each document. To get our heads around this, let’s make a simple pie chart. Copy a row from your document — I’ll choose Row 2 — and paste it into a new Excel sheet.

Now you need to reformat this data a little bit so that Excel can make it into a pie chart. Delete the first two columns, which provide the document number and filename. Then, make one column called Topic and one column called Contribution. Put each topic in the Topic column, and put each topic’s contribution right next to it, in the Contribution column.

The Contribution values may not add up to 1 (meaning 100%), because the TMT is only showing you those contributions beyond a certain threshold. So in the last row, create a topic called Other. For its contribution, type in this function: =1-SUM(B2:B4). (Replace B2 and B4 with the first and last cell numbers of each of your contributions.)

Once you’ve got a grid that looks similar to the one above, highlight the cell values, and from Excel’s Charts menu, select Pie. Now you have a visual representation of the contribution of each topic to the document.

I can’t do this for each of my documents, because I have more than 3,000 of them. But now I have a better understanding of what each row is telling me.

But what are these topics? (1)

Now we sort of understand that each topic contains multiple topics in different proportions. But what do the numbered topics refer to?

That information is contained in a different document, and this next step requires a lot of toggling back and forth between our TopicsinDocs.csv file and our all_topics.html file.

Start by double-clicking all_topics.html. It should open in a web browser. You’ve got a list of topics (or, more properly, word clusters).

It’s important to realize that these topics are listed in no particular order. But the number of each topic corresponds to the topic’s number in your TopicsinDocs.csv spreadsheet. It’s also important to realize that there are more words in this topic than we’re being shown. By default the Topic Modeling Tool just shows us the top 10 words associated with each topic.

So the topic modeling algorithm thinks these are meaningful. We have to figure out why.

In the document we were looking at in the previous step, we saw that Topic 9 made the most prominent contribution. So I’ll click on that topic — here, it’s uk ac london kcl www king http college research centre — and see what it’s all about.

But what are these topics? (2)

When we click on an individual topic, we get a webpage that shows us another way of looking at it: a list telling us which documents feature that particular document most prominently. Here, we see that the file called 3276.txt contains the most words associated with this particular topic.

But what are these topics? (3)

We know that we’re interested in Topic 9, and we know that it’s featured most prominently in this list of documents. Now we can try to start figuring out what it refers to. We might start by just taking a guess. I have a basic familiarity with the corpus in question, and my hunch is that this word cluster is associated with King’s College London, and particularly its research activity. To confirm my hunch, I’ll drill down into the top-ranked documents in our Topic list, as shown in the image above.

Unfortunately, my data doesn’t tell me which words are associated with which topic, but by surveying a number of documents, I should be able to either confirm or question my hunch.

Name your topics

I find that it’s a useful exercise to try to name the topics in my list. Doing so requires me to alternate between reading individual documents and looking for patterns, and it’s an interesting way to look for clusters of meaning that surprise or confuse me. When a topic doesn’t make sense to me, it’s a good excuse to investigate!

If I’m able to name my topics, I’ll have a quick-and-dirty concordance of sorts to the Humanist emails.

By now it should be abundantly clear that no part of this process is “scientific”; it’s just one way of getting your head around a large body of text. So there’s no right or wrong topic name, just schemas that do and don’t help you find interesting features of the text you’re looking at.

Look for patterns

Topic modeling is generally very useful for, say, learning about change over time. And if you’re running the TMT on a small enough set of documents, you might be able to glean that information from your spreadsheet. But if you, like me, have a bunch of documents, it’s quite a bit harder for a human being to suss out patterns without visualization tools.

But you can try. Using Excel’s Sort and Filter functions, you can, for example, show only those documents for which Topic 9 makes the greatest contribution. You can look for documents that seem to have a similar distribution of topics. It’s challenging, but it’s worth doing, if only to get a better sense of how topic modeling works.

Alter variables and try again

The results of any topic modeling test will change a great deal depending on some important decisions that you make. For one thing, the way you divide up documents makes a big difference to your results. In the preceding example, I chose to divide the Humanist emails into 3,584 separate documents. So the TMT is determining topics by looking for clusters in each individual email.

But instead of thousands of individual emails, I could have chosen to feed the TMT five separate chunks of emails, one chunk per year. In that case, the TMT would look for the clusters of words that characterize each year of emails. The image above is my TopicsinDocs spreadsheet for this scenario. You can see that it’s easier to discern gross patterns, but I lose a lot of the nuance I got in the preceding example. And why should a year be the interval at stake here? Might there be a more meaningful way to divide time — perhaps a series of events that I think might be watershed moments for DH? But then again, would that division just tend to confirm my bias?

In the scenario I described here, I chose to ask the TMT for 50 topics. But I could have chosen to ask for fewer, which would tend to make each topic more broad, or more, which would tend to make each topic more specific.

Which set of variables is best? I’m not sure. It depends on what interests you. For me at the moment, what’s “best” is trying multiple strategies, comparing them to each other, and making a list of things that surprise or confuse me.