Here you see a page from the Poetess Archive, a page in which a Visualization Tool has been projected. It is a tool I have been building with Ira Greenberg, digital artist and author of the book Processing which is about the java-based language for creating art originally developed by Casey Rheas and Ben Fry at MIT’s Computation and Aesthetics Lab with John Maeda. I want to give you here a little history of my work with Ira and then later with Professor Jerry Gannod, software engineer and Professor in Miami University’s Computer Science department. I had thought that an analytic visualization tool for assessing data in the poetess archive – a prototype imagined here – should be our goal. I was imagining that one would get a picture like this upon asking the database a research question such as, “How many poems about melancholy were published in literary annuals costing under 2 pounds published were published within 20 miles of London between 1825 and 1835?” Then, I imagined, you could click on one of these little balls, “drill down,” as they say in programming lingo, to the smaller bits of information, and a little box would pop up saying, Samuel Taylor Coleridge, “Work without Hope” in William Ackerman’s The Bijou of 1828.
I worked with my database designer and Ira Greenberg on this for over a year, at which point I finally asked Ira, why do you keep losing all the information that I send you? An artist in both traditional media and the digital, Greenberg clearly did not want to make this tool. “It’s ugly,” he said. And it is. After I asked him what HE wanted to do, we began visualizing poems. What followed were a series of visualizations that Greenberg created of the poem by Felicia Hemans, “Domestic Affections” – we presented these visualizations at MLA last year. First, Greenberg developed a color wheel visualization based on positive (red), negative (blue), and neutral (gray) terms. Concentric rings moving out represent each line in the poem from beginning to end. Here we see its emotional ups and downs, how they cluster, contrast, compare. It is interesting to see that our perception of the color gray or neutrality is affected by the colors surrounding it, just as a peaceful moment after being almost hit by a bus differs from a peaceful moment when one wakes up after pleasant dreams. One sees here not only where positive terms dominate and where sadness prevails, but how such alterations affect one another.
Another visualization crated by Greenberg is what I call an elocutionary diagram. What you see here is of course a still from an animated visualization. Here a worm-like creature crawls through the poem – through “Domestic Affections” – pausing at each punctuation mark for an amount of time determined by the mark’s relative punctuating force. A comma generates a curvature in the arthropod’s body, an exclamation point causes it to stop for a longer period of time and so its tail crashes into its head generating a series of large V shapes. Here we see, in other words, the poem’s elocutionary force pictured as movement through time. Why do Vs cluster and clatter here, why gently rolling comma-lulls there?
And finally, my favorite, is a word fountain. Based on word count, this dynamic picture alphabetizes the words and shoots up water: the higher the water, the greater number of times that word appears in the poem. One immediately becomes thrilled by the spurts, until one realizes that the highest water spurts correspond to words such as “the,” “and,” and “a. Usually the first thing seen in a visualization are errors in data collection, and we neglected to omit stop words here. Leaving aside the data error, we are really interested in this middle range. If one scrolls over it, the word that often recurs appears on top. As is visible here, in “poetess” poetry, “death” is prominent: “Domestic Affections” describes a woman’s desire to have her genius recognized as such, allegedly with fatal consequence. “Death” vies with Depth here. Walt Whitman’s “Out of the Cradle Endlessly Rocking” connects psychological depth with a death song sung by a bird who has lost a lover, its algorithm resembling this one and thereby suggesting that it too is a poetessy poem,
I come now to our new tool, the last version that Ira sent to me. We had decided to code metrical, tropological, and sonic systems in Keats’s “Ode on a Grecian Urn.” In this last version, we picture the meter of the poem, an abstraction that is only possible by putting the visual in motion, which is really Processing’s forte. If we code many poems, we will be able compare poems JUST in terms of meter, all semantic content temporarily laid aside. What poems will be most similar? This is precisely an “emergent truth” called for by Martin Wattenberg (see http://unixgen.muohio.edu/~poetess/vmodel/vmodel.html).
This message from Ira gives the URLs for the various Viz Tools and explains the latest one. All Viz Tools are also available at http://miamichat.wordpress.com
But the goal for the final visualization tool had not been simply to display meter but to abstract from the poem all kinds of literary codes, as Roland Barthes calls them in S/Z: syntactic, tropological, thematic, sonic, and rythmic. I had a really hard time coding the urn for all these systems at once, as you can see here.
As Jerome McGann has said many times in deploring the use of xml for literary texts, it is too hierarchical, and the hierarchies cannot overlap. I had to arbitrarily decide whether assonance would take precedence over consonance, and woe to me should a metaphor fall outside one metrical foot. Both Ira Greenberg and Jerry Gannod encouraged me to create numerous xml encoding schemes rather than just one, and to create them as extensions of TEI rather than rogue xml files.
Partnering with Jerry Gannod and a computer science student doing research on ontologies, I created schemas for and encoded four versions of Keats’s “Ode on a Grecian Urn.” Here you can see the metrical markup scheme, declared in the tei Header.
And here the poem marked up according to that scheme.
This is a version of the Ode coded for sound, marking up instances of alliteration, assonance, consonance, and rhyme.
This describes the categories used to mark up the poem’s syntax.
Sentences can run over several lines and then end mid-verse line, so line tags have to be taken out for the sake of preserving XML’s hierarchical structure.
Here one can see the tropology Ode, as I call it, the Ode to a Grecian Urn marked up for figures – metaphor, synecdoche, chiasmus – images, symbols, ambiguities, and themes.
Here follows the schemas for the odes showing their tree structure, their hierarchies.
There are several things we can do with these codes in addition to picturing them via processing. We can run diff tools, tools that mark out differences in a document, to see how the differently-marked up versions compare. There may be hotspots in the poem that always change, and we might be able to visualize the changes as well.
We know what poems look like in print, but they can be looked at and sounded anew, plumbed for meaning that a single, uni-modal medium leaves latent.
Please feel free to look at my poems and their schemas. I am currently working on creating a schema compatible with TEI P5, work to be completed at the advanced TEI seminar at UCSB this September 14-16, 2009. Those full schemas and the coded Odes will be available after that date, also at the above URL.s