Some results from the Bloomsday hack session, where we discussed small digital Joycean projects we might take on that day and continue working on over the following weeks:

We decided to create a dataset that could be used in Gephi to make something both informative and pretty: a log of the social interactions among characters that could be turned into a social network visualization. Chad Rutkowski read through the Wandering Rocks episode and logged a list of character interactions, which I then turned into a dataset and manipulated in Gephi to produce (click image for larger version):

Character nodes are weighted by the number of edges touching them (i.e. by how many interactions with other people a given character has), so unsurprisingly for anyone who’s read Ulysses, Father Conmee appears as one of the most connected characters in the episode.

Our next step will be to answer some questions about types of character interactions and include these answers in our dataset:

• Do we want to log the direction of an interaction to cover cases where, for example, and conversation is one-sided? (Yes, but this means creating two edges for every dialogue: Bloom > Molly as well as Molly > Bloom).
• What counts as an interaction? (Telegrams, letters, overheard shouts in the street?)
• How do we handle time? (Is Bloom > Molly recorded only one time in our Ulysses datatset, or every time they interact? If the latter, how do we decide when an interaction counts as ended?) If we can find a satisfactory and non-insanity causing way of coding this, we could create a time-lapse visualization of interactions in the novel, perhaps with some sort of cumulative or heatmapping feature.
• How to handle different types of interactions? We discussed assigning different numbered weights to interaction edges so that its easy to see the degree of interaction taking place (was a character imagining a conversation with Bloom, or actually talking to him?), but there’s some difficulty in deciding what types of interaction are deeper than others in order to place these on a spectrum and make visual apprehension easier.

Ben Schmidt also did some neat and quick work with the Circe episode, running a script to gather character names by grabbing the all-caps words in that section and mapping interactions stepwise by linking the names that appears next after a given character in the text.

We’ll continue working on this project as we have time, so if you’re interested in helping out, send us a tweet! The work involved is pretty easy: identifying a section of the novel you wish to attack, then making a list of the characters who interact and ID’ing the type of interaction according to a scheme we’re using.

Cross-posted from LiteratureGeek.com.

Saturday is Bloomsday! It strikes me that we might put our DH nerdiness to more direct Joycean use than simply deciding which Ulysses t-shirts to wear. I propose using this post space to discuss possible small-scale, collaborative Joyce-celebratory projects we might undertake during free time throughout the weekend (not a session necessarily, but a backchannel collaboration). These could be performative (e.g. recording readings of favorite passages? maybe a choral reading? making a small game?) or investigative…

One possibility: I’ve been playing around with the free visualization tool Gephi recently; we might create a simple dataset representing one-to-one character interactions in Ulysses or part of Ulysses–who interacts with whom? How many degrees of separation does the most removed character have from Bloom?–and drop it into Gephi to create a pretty and useful visualization. This would require

1) People signing up to list all the character interactions for a given section in Ulysses (if we only have a few people, we could just divvy up pages in a single rich episode like Circe or Wandering Rocks). For ease of collaboration, it probably makes sense to use the Project Gutenberg e-text (unless the rumored new digital edition drops in time for us to use!).

2) Defining what an “interaction” in Ulysses entails (dialogue? glimpsing someone? thinking about someone?). Or are there other factors we might want to model with a visualization?

3) Creating a basic spreadsheet with two columns: whenever an interaction happens, create a row with Person A on column 1 and Person B in column 2. Unless we decide on doing a one-way directed sort of interaction–e.g. Bloom thinks about Molly doesn’t mean Molly thinks about Bloom at the same time–it doesn’t matter which person in a pair of interacting characters goes in which column. We might also consider adding a “weight” column that keys different weight numbers to “degree of interaction” (e.g. degree of 1 indicates thinking about someone, 2 indicates glimpsing but not being seen, 3 indicates dialogue). I can post a link to a Google Spreadsheet with example rows if people are interested; you might also check out the Gephi sample datasets looking at character interactions in Les Miserables and the Marvel comic universe.

Interested? Or have any other Joycean ideas?