Mining Twitter for Suicide Prevention Amayas ABBOUTE 1 , Yasser BOUDJERIOU 1 , Gilles ENTRINGER 1 , J´ erˆ ome AZ ´ E 1 , Sandra BRINGAY 1,2 , and Pascal PONCELET 1 1-LIRMM UMR 5506, CNRS, University of Montpellier 2 bringay,aze,poncelet@lirmm.fr, 2-AMIS, University of Montpellier 3 Abstract. Automatically detect suicidal people in social networks is a real social issue. In France, suicide attempt is an economic burden with strong socio-economic consequences. In this paper, we describe a complete process to automatically collect suspect tweets according to a vocabulary of topics suicidal persons are used to talk. We automatically capture tweets indicating suicidal risky behaviour based on simple classi- fication methods. An interface for psychiatrists has been implemented to enable them to consult suspect tweets and profiles associated with these tweets. The method has been validated on real datasets. The early feed- back of psychiatrists is encouraging and allow to consider a personalised response according to the estimated level of risk. Keywords: Classification, Suicide, Tweets. 1 Introduction et motivations According to the French website Sante.gouv.fr 1 , nearly 10,500 people die each year in France by suicide (3 times more than traffic accidents). Approximately 220,000 suicide attempts are supported by Emergency department. The eco- nomic burden of suicide is estimated at 5 billion euros for 2009 in France. Sui- cide is a major public health issue with strong socio-economic consequences. The main objective of this study is to detect, as early as possible, people with suicidal risky behaviour. To do this, we focus on recent information retrieval techniques to identify relevant information in texts from the Twitter social network. These messages are used to learn a predictive model of suicide risk. Societal benefits associated with such a tool are numerous. The semi-automatic detection model of suicidal profiles can be used by social web services providers. For example, moderators can use such a model to prevent suicide attempts: by communicating directly with the concerned person, by contacting relatives when possible or by displaying targeted advertisements such as SOS Amiti´ e (trans- lation: SOS Friendship) advertisement which appears when users enter special terms in google search. A detailed analysis of identified messages can also help psychiatrists to identify emerging causal chains between socio-economic inequal- ities and different suicidal practices. 1 http://www.sante.gouv.fr/