Hostname: page-component-76fb5796d-x4r87 Total loading time: 0 Render date: 2024-04-27T11:13:28.487Z Has data issue: false hasContentIssue false
  • English
  • Français

Talking suicide on Twitter: Linguistic style and language processes of suicide-related posts

Published online by Cambridge University Press:  23 March 2020

B. O’Dea ,
M. Larsen ,
P. Batterham ,
A. Calear  and
H. Christensen
B. O’Dea*
Affiliation:
University of New South Wales, Black Dog Institute, Sydney, Australia
M. Larsen
Affiliation:
University of New South Wales, Black Dog Institute, Sydney, Australia
P. Batterham
Affiliation:
The Australian National University, National Institute for Mental Health Research, Canberra, Australia
A. Calear
Affiliation:
The Australian National University, National Institute for Mental Health Research, Canberra, Australia
H. Christensen
Affiliation:
University of New South Wales, Black Dog Institute, Sydney, Australia
*
*Corresponding author.

Abstract

Core share and HTML view are not available for this content. However, as you have access to this content, a full PDF is available via the ‘Save PDF’ action button.
Introduction

Suicide is a leading cause of death worldwide and is largely preventable. The social media site Twitter is used by individuals to express suicidal intentions. It is not yet feasible to contact each Twitter user to confirm risk. Instead, it may be possible to validate risk by linguistic analysis. Psychological linguistic theory suggests that language is a reliable way of measuring people's internal thoughts and emotions; however, the linguistics of suicidality on Twitter is yet to be fully explored.

Objectives & aim

The aim of this study is to characterise the linguistic styles of suicide-related posts on Twitter for the purposes of predicting suicide risk.

Methods

The Linguistic Inquiry and Word Count (LIWC) program was used to compare the linguistic features of suicide-related tweets previously coded for suicide risk by humans with a set of matched controls. Logistic regression was then used for predictive modelling.

Results

The suicide-related tweets had significantly different linguistic profiles to the control tweets. The “strongly concerning” suicide tweets were found to have fewer words than all other tweets and not surprisingly, references to ‘death’ were significantly higher in this group. A number of other results were found. The final model which distinguished “strongly concerning” suicide risk from the controls was found to have 97.7% sensitivity and 99.8% specificity.

Conclusions

This study confirms that the linguistic features of suicide-related Twitter posts are different from general Twitter posts and that these linguistic profiles may be used to predict suicide risk in Twitter users.

Disclosure of interest

The authors have not supplied their declaration of competing interest.

Type
EW609
Information
European Psychiatry , Volume 33 , Issue S1: Abstracts of the 24th European Congress of Psychiatry , March 2016 , pp. s274
Copyright
Copyright © European Psychiatric Association 2014
Submit a response

Comments

No Comments have been published for this article.