Italian author and meta-fabelist Italo Calvino arranged a grid of tarot cards as an engine to propel his 1973 novel The Castle of Crossed Destinies. The interwoven lines and columns of this grid were used by a narrator to reconstruct the stories of a whimsical cast of characters who had become mute, traumatized after venturing through the wilderness. These kind of narrative games are not the exclusive domain of authors, as passionate readers—or viewers–can bring enormous energy and insight into revealing the structure, connections and veiled references that are baked into a sufficiently intricate plot. Buenos Aires-based information visualization specialist Santiago Ortiz’s most recent project has exactly these ambitions, and provides a suite of interfaces for exploring the scripts and character relationships across the 121 episode corpus of the ABC television series Lost.
Working with data scraped from the popular fan site Lostpedia, Lostalgic opens to a list-like super-script of every line of dialogue from the six seasons of the Lost. Zooming down to the resolution of a single episode reveals a staccato arrangement of bar graphs that represent each scene and exchange that occurs between characters. Cells can be hovered over to reveal individual lines and the interface provides a chronological map of the series. Clicking through to the matrix view switches over to more dynamic network of character relationships and this is where the utility of Lostalgic really becomes apparent. The matrix visualization displays shifting social constellations – how characters interact within an episode or across the entire series. Opening up the link between two characters reveals a display with a venn and bar diagrams that maps their shared screen time and clicking on an individual provides tree maps that situate the character in relation to the rest of the cast. The matrix repurposes many of these same visual tropes and represents each episode as a narrative grid for mapping character interactions. There is a also a renactment view that provides a chronological retelling of the series as a ‘thumbnail slideshow’.
Lostalgic is a successful undertaking because it completely ignores Lost‘s mythology and its confounding plotlines (and plotholes!) and instead focuses on depicting the underlying structural narrative and how it emerges from a sprawling ensemble cast. Ortiz is astute in observing that Lost surfed a big “zeitgeist wave” to mass popularity and part of the reason that the show was so habit forming for audiences was that it echoed the ascent of “social networked narratives” during its 2004-10 run. Less pragmatically and more poetically, it repurposes the 44 minute television drama format as raw material for the construction of a combinatorial playground that demands exploration. This is an important project that invites comparison to Frederic Brodbeck’s Cinemetrics (2011) and more archaic methods of shot length tracking, but instead of yielding an ‘aesthetic DNA’ of colour and composition, Lostalgic finds its rhythm and form within a much more expansive dataset and through an extremely close reading of the script.
CAN posed a related line of questioning to Santiago Ortiz about his new project – note his expansive commentary below.
I think the most obvious place to start would be in asking why did you select Lost as the basis of this script-mapping experiment? Did the desire to develop these interfaces emerge from your fascination with/fandom of the series or was your selecting it a pragmatic decision based on the amount of data available at Lostpedia?
Both! Being a big fan of Lost I naturally ended up at Lostpedia, and in the very moment I discovered that all scripts were there, available and, at first glance, coherently formatted, I knew I would create something with that material.
Actually, when I was eagerly consuming the first season, I started fantasizing about a visualization project. I’ve been always interested in network visualization and it was clear Lost is based on a networked narrative, more accurately a social networked narrative. When the first season was being broadcast, it was also a time in which Facebook started being very well known and the concept of social network was entering popular culture. That is, no doubt, part of the success of the series: it surfed a big zeitgeist wave.
Previous to Lost you had choral TV shows and movies, based on multiple characters and intersecting stories, but none systematically explored almost all possible differentiated characters relations like Lost did. That’s in its DNA, it’s basically how they wrote the scripts, by combining groups of characters, inventing narrative excuses to place different groups of characters in different parts (or times!) in the island, like if they were bacteria placed in different petri dishes exposed to differentiated environments. Kate+Sawyer in a cage, Locke+Boone in a crashed plane, Locke+Desmond on a hatch… a big etcetera, and that’s only for pairs. Not that [Lost writer] Damon Lindelof told me that! Not that I needed to be told either. If you want to ‘feel’ these combinatorics going on, just open Lostalgic in matrix mode, and press the ↓ key several times, you’ll see how the relations patterns change dramatically episode to episode.
At that time of the first season ABC published its own network interactive visualization, adequately called ‘connections’ (it seems that is not longer available), it was pretty well done and that probably deter me to try my own approach. That, and not having any data.
The very day I found the scripts in Lostpedia I started working.
It’s interesting that this social networked narrative was created by small social collaborative network (more than 35 writers), and then comprehensively documented by a community (Lostpedia contains more than 7000 articles), another collaborative social network. And then, across the internet, we have seen a lot of other spontaneous projects created out of the TV show (maps of the island, games, comics, conspiracy theories, visualizations, etc.), mine is just one to add to the list. This one for instance, is fantastic:
However, I do think this is potentially a general tool that could be used to visualize, analyze and enjoy any other TV show script, and eventually other narrative texts, in a similar way this great project does with movies.
Can you describe some of the development challenges you faced with this project?
Grabbing the data could have been the hard part in this project but it wasn’t. I was gladly surprised by how consistent was the format throughout the 115 episodes… with very few exceptions for which I had to write some contingent code or even perform some changes by hand.
The really difficult part was to deal with the hierarchical structure, to define the statistics that works in different levels in the hierarchy and to develop the API aimed to respond to all the possible queries. For instance, for a single character I need to know how many times he or she speaks in a scene, in how many lines of an episode they appear, in how many acts, in how many episodes in the entire series… and for two characters I need to know the co-occurrences in the series, per episode, per line, etc.
In terms of visualization each view had its own challenge. For the matrix I wanted to do something different… not only to express the weight of the relation for each pair of characters but to say something else. I ended up using proportional surface Venn diagrams but instead of drawing the entire diagram I just drew the intersections, the lenses and this works because the lens by itself conveys the complete information of the entire Venn diagram (a portion of circle, no matters how small it is, completely defines the circle)! Not that it is easy for the eye-brain to complete the diagram, but it’s a matter of training (the eye-brain system is very good completing information gaps). The name ‘lens’ for this visualization method is not merely a metaphor, this shape actually describes the section of an optic lens whose optical behaviour is determined by the radius and distance of the two circles.
You mention that one of the goals for the project is to allow users to “read and enjoy the series in a different way” – could you elaborate on this intention?
I believe books, movies and in general stories could be visualized in ways persons not only will learn about the contents, the context and the structure of the narrative but will actually read in a different ways the story, or, if you want, will read another story out of the atoms and molecules of the allegedly analyzed one (and I use the word ‘read’ in the most wide hermeneutical possible sense). These aren’t new ideas at all, for many that’s exactly what literature and art criticism should do: build new meaning out of the previously existing one. When it comes to create interactive visualization, or, in general, interactive creation based on pre-existent narrative material, I think there are multiple unexplored ways to create new meaning, new stories… or to re-tell the same story (which is as impossible as to take a bath twice in the same river, as Borges perfectly explained in his Pierre Menard, Author of Quixote story). ‘Re-telling’ has been explored in digital arts but not so much in visualization.
When I was building the first view (the big zoomable index) I found myself reading the episodes, line by line… reviving the series, Lost in nostalgia (the name comes from this)… thus I decided to create a view aimed to reenact the episodes. I’ve already seen a few episodes with the reenactment view and it works pretty well. Again, as happened with the lenses visualization, our brain is good at completing incomplete streams of information, so with just character images, dialogues and timing, it’s possible to build a theatrical experience. Scott McCloud has written on how important the non-existent information between frames is and how our brain ‘fill the gaps’ with narrative substance.
Ok, so we’ve spoke in generalities about how this project might re-present the narrative of Lost, how has executing this project changed your thinking about the series?
Mainly I’ve learnt about the way the scripts are written, about techniques the writers used. As I commented before, it was clear to me that the writers actually were looking for multiple characters combinations. Other interesting patterns that can be seen with the graph or the matrix views are related with the rhythms, how from one episode to the following the characters network changes abruptly. In general a similar characters network reappears two or three episodes later (meaning that the particular story regarding these characters is resumed). The alternation of stories allows to develop more cliffhangers and Lost is full of them.
Going through episodes in the matrix mode another thing became clear to me and that is the importance of the changes in balance in networks. Some episode are very well balanced, meaning that the number of appearances for each character is similar, and the co-occurrences as well. This is the case of “Walkabout” (episode 4) and also the case of the final episode, “The End”, in which all characters and relations are treated as equal, on purpose. Two episodes before there’s “Across the Sea”, the most eccentric and unbalanced episode in the series. First and, overall, last episodes in each season tend to be very balanced.
I’m not completely satisfied with Lostalgic in his current version as a tool to see and explore global patterns. I plan to add a couple of views that I’ll be more directed towards that goal, in particularly one that will depict a network of episodes (instead of characters), whose relations will be defined by networks similarity (so the result will be a network of networks). In this view things such as the alternation of balanced and unbalanced episodes will be more clear. I also want to visualize the writers involved in each episode.