BACKGROUND: Suicide is a leading cause of death worldwide. Identifying those at risk and delivering timely interventions is challenging. Social media site Twitter is used to express suicidality. Automated linguistic analysis of suicide-related posts may help to differentiate those who require support or intervention from those who do not. AIMS: This study aims to characterize the linguistic profiles of suicide-related Twitter posts. METHOD: Using a dataset of suicide-related Twitter posts previously coded for suicide risk by experts, Linguistic Inquiry and Word Count (LIWC) and regression analyses were conducted to determine differences in linguistic profiles. RESULTS: When compared with matched non-suicide-related Twitter posts, strongly concerning suicide-related posts were characterized by a higher word count, increased use of first-person pronouns, and more references to death. When compared with safe-to-ignore suicide-related posts, strongly concerning suicide-related posts were characterized by increased use of first-person pronouns, greater anger, and increased focus on the present. Other differences were found. LIMITATIONS: The predictive validity of the identified features needs further testing before these results can be used for interventional purposes. CONCLUSION: This study demonstrates that strongly concerning suicide-related Twitter posts have unique linguistic profiles. The examination of Twitter data for the presence of such features may help to validate online risk assessments and determine those in need of further support or intervention.
BACKGROUND: Suicide is a leading cause of death worldwide. Identifying those at risk and delivering timely interventions is challenging. Social media site Twitter is used to express suicidality. Automated linguistic analysis of suicide-related posts may help to differentiate those who require support or intervention from those who do not. AIMS: This study aims to characterize the linguistic profiles of suicide-related Twitter posts. METHOD: Using a dataset of suicide-related Twitter posts previously coded for suicide risk by experts, Linguistic Inquiry and Word Count (LIWC) and regression analyses were conducted to determine differences in linguistic profiles. RESULTS: When compared with matched non-suicide-related Twitter posts, strongly concerning suicide-related posts were characterized by a higher word count, increased use of first-person pronouns, and more references to death. When compared with safe-to-ignore suicide-related posts, strongly concerning suicide-related posts were characterized by increased use of first-person pronouns, greater anger, and increased focus on the present. Other differences were found. LIMITATIONS: The predictive validity of the identified features needs further testing before these results can be used for interventional purposes. CONCLUSION: This study demonstrates that strongly concerning suicide-related Twitter posts have unique linguistic profiles. The examination of Twitter data for the presence of such features may help to validate online risk assessments and determine those in need of further support or intervention.
Entities:
Keywords:
Twitter; linguistic analysis; prevention; social media; suicide
Authors: John Torous; Mark E Larsen; Colin Depp; Theodore D Cosco; Ian Barnett; Matthew K Nock; Joe Firth Journal: Curr Psychiatry Rep Date: 2018-06-28 Impact factor: 5.285
Authors: Dan J Stein; Steven J Shoptaw; Daniel V Vigo; Crick Lund; Pim Cuijpers; Jason Bantjes; Norman Sartorius; Mario Maj Journal: World Psychiatry Date: 2022-10 Impact factor: 79.683