Interactive Installation & Data Sculpture
Based on approx. 12.5 million Twitter messages which were aggregated in real-time, emoto captured trending topics and how they were discussed online in an interactive visualisation which was running live in parallel to the Games in July and August 2012. Each Tweet was annotated with a sentiment score by the project’s infrastructure using software provided by Lexalytics. This data formed the basis for an extensive profiling of London2012 which was finally documented in this interactive installation and physical data sculpture at the WE PLAY closing exhibition of the Cultural Olympiad in the Northwest.
In the final installation we have focused on presenting all data emoto has gathered and the insight we have derived from it.
The core of the installation is a physical data sculpture consisting of 17 objects, each representing all Tweets we have collected during one day of the Olympics. Mapped onto this physical sculpture we have then projected individual heat maps for the most interesting themes we have identified while observing emoto during the Games. Users were able to navigate through these themes using an interactive controller and thus explore our archive.
The second key element in the installation was the 9.50 meter long Sentigraph representing giving a high-level overview over London2012 as seen by emoto.
Together with Moritz Stefaner we have been responsible for the conceptual direction, design and technical development for the entire project in close collaboration with Drew Hemment from FutureEverything.
The emoto data sculpture represents message volumes, aggregated per hour and sentiment level in horizontal bands which move up and down according to the current number of Tweets at each time. This resulted in simplified 3-dimensional surfaces which allows visitors to identify patterns in message frequency distribution more easily. And while not being specifically designed in this direction, the surfaces also nicely support haptic exploration.
The Sentigraph is a 9.50 meter long multi-layered print, designed for the visitor to explore the overall timeline of the Olympics. Th graph shows the average mood for all events and topics as tracked by emoto. It was printed on transparent acrylic glass and offset from the wall by approx. 7cm to reveal the content behind it. All messages were directly attached to the wall and have been selected for peaks in the graph based on the occurrences if the Tweet text.
From the emoto archive, we have aggregated frequencies of messages per hour and sentiment level into 2-dimensional heat maps. These heat maps were then transformed into 3D geometry and finally CNC-milled in collaboration with our manufacturer Tischlerei Bächer using Polyurethane-foam (’Chemiwood’). Additionally, the objects were painted using a dual component paint with particles to optimise the surface for projection.
On top of this sculpture we have then projected multiple heat maps, only displaying events for the currently selected theme (i.e. Team GB). A visitor could control which theme to show using a Griffin Powermate. Pressing the button would cycle through the themes. Rotating it would move the cursor along the timeline, showing most retweeted messages for each hour and theme.
The projection mapping was custom developed in Processing as part of the installation software. The 2D heat-maps were generated in Tableau and used as textures for the mapped virtual geometry. The final outcome for these textures was designed in multiple quick iterations exploring the use of many geometric shapes for the heat maps. The colours have been chosen based on the overall emoto colour scheme but were optimised for projection onto dark grey material.
We have also decided to expand the emoto sentiment scales from +-6 to +-12 levels, since working with physical material on a 3m long table and object allowed us to display the information we have more granularly.
The displayed Tweets were determined in a Map/Reduce workflow using the emoto MongoDB archive. This enabled us to work more flexibly with the 50GBs of data. The final principle counts occurrences of Tweet texts for each hour and selects the most occurring Tweet for display. While this mechanic provided very nice results, it could be further improved to by additionally limiting the aggregation to one sentiment level at a time and including edit distance in the text comparison process.
Infrastructure design & development by Gerrit Kaiser.
Citizen journalism by Andy Miah
Evaluation and blogging by Ege Sezen.
Project management by Leon Seth and Nick Lawrenson.
Communications by Jo Williams, Anita Morris Associates.
Emoto is produced by FutureEverything and Studio NAND.