Online Suicide Identification in the Framework of Rhetorical Structure Theory (RST)

Abstract

Background: Suicide is a serious social problem. Substantial efforts have been made to prevent suicide for many decades. The internet has become an important arena for suicide prevention and intervention. However, to the best of our knowledge, only one study has analyzed suicidal comments online from the perspective of rhetorical structure with incomplete rhetorical relations. We aimed to examine the rhetorical differences between Chinese social media users who died by suicide and those without suicidal ideation. Methods: The posts of 15 users who died by suicide and 15 not suffering from suicide ideation were annotated by five postgraduates with expertise in analyzing suicidal posts based on rhetorical structure theory (RST). Group differences were compared via a chi-square test. Results: Results showed that users who died by suicide posted significantly more posts and used more rhetorical relations. Moreover, the two groups displayed significant differences in 17 out of 23 rhetorical relations. Limitations: Because this study is largely exploratory and tentative, caution should be taken in generalizing our findings. Conclusions: Our results expand the methods of RST to the online suicidal identification field. There are implications for population-based suicide prevention by combining rhetorical structures with context analysis.