Sentiment analysis and mood trees

An agency called Kerve designed what they called a mood tree to capture the sentiment of the attendees of a design festival using every tweet containing the event hashtag from a Twitter API and identified the most meaningful word of each. The algorithms were based on Stanford's natural language parser. The audience could change the colour of the tree via their web app, which asked them to choose the colour that best suited their mood. We would have to find a way of securely and robustly storing a Mac Mini or its equivalent into the base of the tree. As I wrote in my individual blog, this is why it might make sense to have a tree that is either entirely man made or a combination of natural and sculptural (my preferred solution).

The Kerve Mood tree

They described the technology as: The data was routed through a remote web-socket server, which controlled communication to the node.js server running on a Mac Mini within the base of the installation. This server communicated directly to four master Arduinos via the serial port, each of which communicates to a further six slave Arduinos.

Each of the 24 total slave Arduinos is wired into one of the laser cut lanterns suspended from the tree, via the chrome pipes rising from the centre. Each lantern contains nine LED matrices, making up the scrolling text display, and two colour changing LED strips, giving everything a bright and colourful ambience.

Every new tweet containing the event hashtag scrolled through one of the matrices, and the most meaningful word was then highlighted.