PHASE: Learning Emotional Phase-aware Representations for Suicide Ideation Detection on Social Media

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics, pages 2415–2428 April 19 - 23, 2021. ©2021 Association for Computational Linguistics 2415 PHASE: Learning Emotional Phase-aware Representations for Suicide Ideation Detection on Social Media Ramit Sawhney∗ IIIT Delhi ramits@iiitd.ac.in Harshit Joshi∗ University of Delhi harshit113@ducic.ac.in Lucie Flek Mainz University of Applied Sciences lucie.flek@hs-mainz.de Rajiv Ratn Shah IIIT Delhi rajivratn@iiitd.ac.in Abstract Recent psychological studies indicate that in- dividuals exhibiting suicidal ideation increas- ingly turn to social media rather than mental health practitioners. Contextualizing the build- up of such ideation is critical for the identifi- cation of users at risk. In this work, we fo- cus on identifying suicidal intent in tweets by augmenting linguistic models with emotional phases modeled from users’ historical context. We propose PHASE, a time-and phase-aware framework that adaptively learns features from a user’s historical emotional spectrum on Twit- ter for preliminary screening of suicidal risk. Building on clinical studies, PHASE learns phase-like progressions in users’ historical Plutchik-wheel-based emotions to contextual- ize suicidal intent. While outperforming state- of-the-art methods, we show the utility of tem- poral and phase-based emotional contextual cues for suicide ideation detection. We further discuss practical and ethical considerations.1 1 Introduction Every 10.9 minutes, a person dies of suicide (Dra- peau and McIntosh, 2020). Suicide ranks as the second leading cause of death for 14-35 year- olds (Hedegaard et al., 2020) in US. Extending appropriate clinical and psychological care to sui- cidal people relies on identifying those at risk. Un- fortunately, 80% of patients do not undergo clinical treatment, and about 60% of those who died of sui- cide denied having any suicidal thoughts to mental health practitioners (McHugh et al., 2019; Franklin et al., 2017). In contrast, people exhibiting suicidal ideation often use social media to express their feel- ings (Coppersmith et al., 2014, 2016, 2018; Robin- son et al., 2016; Reger et al., 2020), with eight out ∗ Equal contribution 1Code is available at https://github.com/ midas-research/phase-eacl/ guys work hard it is worth it! btw we also reached the football finals #goforvictory 14 May 2012 1:42 PM i have a good feeling about todays match we will definitely win! world, here we come 15 May 2012 08:28 AM lost my dad in an horrific accident, dropped out of college i feel hopeless just want to stab my chest 25 Aug 2015 12:58 AM free me from all this or i will do something myself 25 Aug 2015 01:17 AM positive emotions negative emotions negative emotions TWEET TO BE ASSESSED Tweet to be assessed 2012 2013 2014 2015 i woke and i dont feel good, nothing to look forward to, i feel a whole lot more tired ~ 3 years 7 June 2015 07:13 AM i am joining therapy from tomorrow, lets hope i feel better after a few sessions 13 June 2015 04:41 AM ~ 2 months Phase 1 Phase 2 Phase 3 A user's tweeting history with timestamp information variation in emotions expressed suicidal intent Figure 1: …

Content truncated to 500 words