Untitled

Abstract

This work aims to infer mental health status from public text for early detection of suicide risk. It contributes to Shared Task A in the 2019 CLPsych workshop by predicting users' suicide risk given posts in the Reddit subforum r/SuicideWatch. We use a convolutional neural network architecture to incorporate LIWC information at the Reddit post level about topics discussed, first-person focus, emotional experience, grammatical choices, and thematic style. In sorting users into one of four risk categories, our best system's macro-averaged F1 score was 0.50 on the withheld test set. The work demonstrates the predictive power of the Linguistic Inquiry and Word Count dictionary, in conjunction with a convolutional network and holistic consideration of each post and user.

Leveraging Reddit for Suicidal Ideation Detection: A Review of Machine Learning and Natural Language Processing Techniques

“…The social stigma related to having suicidal ideations has a particularly significant effect. The fear of social stigma has been shown to discourage individuals at risk of suicide from discussing their experiences in person and seeking support [ 22 , 24 , 25 , 26 ]. Further, it obstructs the extant suicide-risk screening methods, such as questionnaires and interviews, since they require patients to explicitly disclose their intentions to commit suicide [ 27 ].…”

“…According to a meta-analysis of 71 studies, on average, nearly 80% of people in non-psychiatric settings—primary healthcare patients, general population, military personnel, and incarcerated individuals—who died by suicide did not reveal their suicidal intentions when they were surveyed before their suicide attempt [ 28 ]. Thus, there is a need for novel suicidality detection methods that do not require face-to-face interactions [ 24 ]. In this case, detecting suicidal ideations on online platforms can be more effective since the anonymity of social media and forums enables people to openly share their struggles with suicidal thoughts without fear of judgment [ 11 , 16 , 29 , 30 ].…”

“…For example, if a Reddit post with 100 words is analyzed with LIWC and 10 words relate to negative emotions, the post scores 10% in the negative emotion category. Allen et al [ 24 ] used LIWC to extract features from post titles and bodies. The features were used as inputs to the convolutional neural network (CNN) algorithm.…”