Understand Emotions in Suicide Notes

Suicide, the fourth leading cause of death for adults between the ages of 18 and 65 years in US, takes away over 34,000 lives each year [1]. The messages/notes, which are left before people commit suicide, usually reflect the status of their emotions/sentiments. By automatically analyzing these notes, we want to identify their emotions/sentiments, which can be used for monitoring health status of people.

As a preliminary study, upon a collection of 600 suicide notes, we focus on the problem of classifying each sentence into 15 categories: sorrow, information, hopefulness, instructions, thankfulness, forgiveness, love, pride, abuse, anger, happiness, hopelessness, guilt, blame and fear. And we are able to identify patterns indicating different sentiments. For example, the pattern “can not go on” indicates hopelessness in sentence “I really can not go on any more.” And the pattern “take care of” indicates instruction in sentence “Please take care of our children!”

Figure 1 : A Hybrid Classifier for Sentence-level Multi-class Classification of Suicide Notes

As a next step, we plan to apply our research into social networks. We want to capture sentiments of people in social media (e.g., microblogs, facebook). Moreover, we’d like to model how sentiments of people are changing over time. Last but not least, we want to answer the question “what are the reasons behind the change of sentiments?”.

[1] Facts and Figures about Suicide: http://www.afsp.org/index.cfm?fuseaction=home.viewpage&page_id=050fea9f-...

publication

Wenbo Wang, Lu Chen, Ming Tan, Shaojun Wang and Amit P. Sheth. "Discovering Fine-grained Sentiment in Suicide Notes", Discovering Fine-grained Sentiment in Suicide Notes. Biomedical Informatics Insights, 2012 (to appear)