The heart of the project is an algorithm of Sentiment Analysis based on the open-source library developed by U. Krcadinac, P. Pasquier, J. Jovanovic & V. Devedzic. Synesketch (An Open Source Library for Sentence-Based Emotion Recognition, IEEE Transactions on Affective Computing 4(3): 312-325, 2013).

The algorithm classifies emotions into six distinct categories: happiness, sadness, fear, anger, disgust, and amazement. It performs text analysis on each tweet at an approximate rate of 30 tweets per second. This analysis processes the tweet word by word using a comprehensive lexicon of over 5,000 items, each of them having an assigned emotional score based on its semantic meaning. Additionally, heuristic rules are applied during the analysis, such as detecting negations or amplifying scores for capitalised words to reflect their increased significance.

After processing, each tweet is quantified into six distinct values, one for each emotion, ultimately identifying the predominant emotion associated with the tweet. The overall emotional state is then represented by the percentage distribution of each emotion, derived from the aggregate number of tweets corresponding to each emotional category.

The sound component is a crucial aspect of AMYGDALA since it serves as a metaphorical representation of the emotional analysis and recognition process. The sound system is managed through Max MSP, utilising six distinct sound textures corresponding with the six identified emotions. These textures are mixed via the OSC (Open Sound Control) communication protocol, which integrates data derived from the tweet analysis.

During the initial phase of data collection, the system applies distortions, minimal reproduction delays (ranging from 0.1 to 100 milliseconds), and pronounced decay effects, resulting in coarse and indistinct soundscapes. As the emotional analysis progresses and emotions become clearer, the sound evolves to reveal a melody that reflects the emotional percentages obtained.

Additionally, the Max MSP patch is designed for a quadraphonic setup, allowing the six sound textures to rotate around the spectator. This creates a disorienting auditory effect until the emotions are fully identified, concluding the AMYGDALA cycle. Each cycle is distinct, ensuring a unique auditory experience each time the installation is activated.

The light installation of AMYGDALA has been created for the 41 LED bars distributed throughout the two gardens of CUBO. Each column is made up of a matrix of 3,071 high luminosity RGB LEDs, for a total of 125,952 LEDs, each controlled from the server room through 2.75km of optic fibre. The columns were designed to work 24 hours a day, 365 days a year. An internal cooling system allows a constant temperature to be kept, and a flow of air is blown onto the front glass panel to avoid misting.

In a particular section of the garden between the two wings of CUBO buildings, the columns are placed in a circular formation. This area houses a 360° audio system that envelops the spectators within the heart of the installation.