Sosolimited has been using software to perform textual analysis on political debates since 2004, when it first used these visualizations to analyze a debate between John Kerry and George W. Bush. Since that time, the software has been adapted to reinterpret a number of similar broadcasts, most notable in their project ReConstitution, which was performed as an audio visual remix to audiences in three different cities during the debates. The debate between John McCain and Barack Obama was transcribed in real time, and the text was then run through analyses to identify a number of trends. The software was designed to pick out commonly used words, their length, and keywords belonging to certain categories, such as 'you', 'me', and 'the other guy', which recorded who each candidate was addressing in each statement. The data was then visualized to show the audience these patterns in a decipherable and interesting way. Keywords would be highlighted by a colorful block with tags listing their number of occurrences, and large headlines, variably reading 'you', 'me', and 'the other guy', would pop up next to a live stream of the debate, displaying both the number of times each group was addressed and all of the strings of text containing key terms. Aside from performing textual analysis and data visualization, the people controlling the performance also distorted the images and audio coming from the live stream. Faces of the candidates were sometimes blurred by a glitch effect and their voices manipulated to sound as if trying to protect the anonymity of the candidate. All of this was done in a way which preserved the crucial content of the debate, meaning that no audio was so distorted that one could not keep track of what each candidate was saying.
The purpose of this project is to provide a new lens to look at political commentary through. Too often we are stuck listening to repetitive dialogues that rarely reflect what the candidate will actually be able to do given their election. Hollow promises and propaganda seem to be the status quo of political debates today, and Soso has made it their mission to try and mine some tangible information from these broadcasts. Counting the number of times a words has been used, or who each candidate addresses most often, is hardly going to flesh out the true 'meaning' of what is being said, but it provides and interesting commentary on the repetitive and predictable nature of presidential debates.
Filip Visnjic with Creative Applications notes:
"Whether the project creates an insight into what the world leaders “mean” is somewhat debatable. Extracting words out of their context and interpreting facial expressions may only begin suggest new narratives disconnected from their origin. The fact of the matter is that most of these political debates are no more than theatre, created for public and press media to feed on. The truth is that we, without the additional software, create meaning depending on our social standing, education and what might be most relevant to us. Sosolimited’s segmentation nevertheless does provide an insight into how machines interoperate live television but whether we can relate to these machines is altogether another matter. "(Visnjic, Filip (3/7/2010), Sosolimited – reConstitution )
One important thing to note in Visnjic's assessment of ReConstitution is that this performance is not meant to uncover some hidden meaning in the words being analyzed, but more significant in the way that it removes you from your rooted ways of interpretation. Your sociopolitical tendencies often warp the way you interpret the things being said; this piece points out that sometimes you need to look on with a new perspective to see underlying meaning.
Soso works with textual data in a number of other projects. Similarly, in a piece called Set Top Box, Soso uses the software to analyze transcripts of soap opera shows to identify characters emotions, highlighting key words with a color representative of the associated emotion. Again, this piece also includes an aspect of visual distortion as it uses the color and transcript during commercials to generate a 3D landscape in the projection.
Unlike Reconstitution, this piece is mostly aimed at displaying artificial intelligence in an interesting way. There is something striking about hearing a robotic voice accurately saying "sadness detected" when a character gets upset on screen. This piece is striking not because it uncovers hidden patterns in the broadcasts, but because it can pick up on human emotion. On the other hand, Reconstitution is striking primarily due to its ability to pull out patterns that are not so easily identified without the software present.
Another artist who does a significant amount of work with text based visualizations is Charles Sandison. In his piece Rage, Hope, Love, Despair, Sandison visualizes text through a projection that color codes words depending on which emotion it is related to(rage, hope, love, or despair). Sandison programs the words to interact with each other in ways that serve the purpose of the message he is trying to convey.
Unlike, Reconstitution, which is displayed on a theatre like projection, Sandison's work on this project uses all of the buildings surfaces in his projections, giving his text visualizations a greater sense of life as they move through the physical space and interact with one another. Sandison uses physical space and programmed behaviors in order to help breathe life into his words. Where he is using words to display emotional connection between words such as love and hate, Soso is using words in a purely informational way. The text being fed into the software is pulling out hard data, such as word frequency, and visualizing the words based on these quantitative measures. While Sandison bases his visualization on aesthetic choices regarding the complex interconnectivity of emotions, Soso's project comes as one that is purely factual, without any space for artistic interference in the data.