Big data, machine learning e intelligenza artificiale: nuove prospettive per la prevenzione in psichiatria. L’esperienza del progetto InsideOut

  • Federico Fiori Nastro
  • Francesco De Michele


The article describes the setting up and development of an experimental protocol aimed at enhancing a software that can improve our knowledge of the habits, the behaviours and moods of adolescents. For instance, pre-clinical detection of distress and discomfort can favour approaches aimed at tackling the huge clinical and socio-economic burden of mental illnesses in Western countries. The so-called Inside-Out tool can analyse texts on Twitter. This tool has been used within our study to analyse Italian and English tweets written by adolescents aged between 12 and 19 years. Inside-Out is an infrastructure that works based on Computation Linguistics and Machine Learning techniques to gather the meaning and the sense of text messages and to improve the performance of the system through the experience. This process is based on an original model of distress characterization that has been developed by trained psychiatrists to identify different semantic domains (Life Event) with which tweets can be correlated. Each tweet has been associated with a Sentiment (positive or negative) and an Experience (dangerous or help-seeking). The authors illustrate the results and potential of a method that depicts the feelings and the experiences of the young without introducing an external bias, i.e. the observer, and allows to gather important epidemiological data in the community in a very short time.