An across online social networks profile building approach: Application to suicidal ideation detection

With the spreading use of online social networks (OSNs) such as Facebook, Twitter and Youtube, most people find that writing about their feelings and sharing their preferences and thoughts in social media is actually easier then articulating them in real life. Specifically, these networks are increasingly associated with different social phenomena such as diseases, depression and even suicide. In this context, most people who have suicidal ideations and are active in OSNs give signals of their intentions. It is therefore essential to increase efforts working on the detection of profiles whose owners have suicidal ideations as an attempt to prevent suicide. For that purpose, in this paper we propose a new method that automatically detects suicidal users through their created profiles in OSNs. Our contribution consists in considering profiles from multiple data-sources and detecting suicidal users based on their available shared content across OSNs. We extract several types of features from the posting content of users to build a complete profile that contribute to high suicidal user prediction. We employ supervised machine learning techniques to distinguish between suicidal and non-suicidal profiles. Our experiments on a dataset, which consists of persons who had died by suicide, demonstrate the feasibility of identifying user profiles from multiple data-sources in revealing suicidal profiles. (C) 2022 Elsevier B.V. All rights reserved.