Dr Ted Pedersen is the author of 'Rule-based and Lightly Supervised Methods to Predict Emotions in Suicide Notes', recently published in the Biomedical Informatics Insights Supplement.  We asked Dr Pedersen to tell us about the background of his paper.

To start please tell us about the challenge this year.  Why did you decide to become involved, and what goals did you and/or your team expect to accomplish?

I was particularly interested in learning more about the nature of suicide notes. This is something that we only tend to hear about in the case of “famous” or celebrity suicides, and I was very intrigued by the idea of being able to study a larger corpus of notes written by more ordinary people.

My main goal for the challenge was simply to acquire a better understanding of suicide notes, and also see how well some ideas from Natural Language Processing could be applied to this kind of data. This was really more of a personal project for me, and so I worked on this by myself, and spent quite a bit of time reading and studying the notes before working on my actual system.

In writing this paper what were the particular challenges you faced?  How did you overcome these challenges?

Writing the paper was actually not too difficult, since I knew I wanted to mostly focus on what I observed in this data, and how that translated into a system. In my case since I spent a good bit of time studying the suicide notes, my best performing system turned out to be a manually created set of rules. Those rules were based on my own observations and intuitions about the emotions present in suicide notes, and really didn't use much in terms of machine learning or other automated methods.  This is a departure for me, since much of my other work is more oriented towards machine learning and other automated techniques. In fact I have never before published a paper about a manually built rule based system.

What has been the major benefit for you in the work discussed in your article?  How has it contributed to our knowledge of the area?

I think I came away from this with a much better understanding of suicide notes. What I learned is apparently well understood by people who work with suicide, and that is that many suicide notes are actually fairly “matter of fact” and are really not very emotional at all. Rather they reflect someone who has made up their mind about ending their life, and are now often dealing with final instructions and farewells. I had expected the notes to be much more emotional and even tortured, but it seems that many times people who have gotten to the point of writing a suicide note have actually worked through those feelings, and so the resulting note is in some cases quite calm. This is not always true of course, but it was true more often than I would have thought.

To be candid, I'm not sure that my system or this paper makes a really large contribution to the area. I think at best it provides a kind of baseline measure of performance that shows manually created rule based systems can perform quite well, but will often lag methods based on machine learning. However, I am quite certain I learned much more about this domain than if I would have relied on machine learning or other automated methods. In a sense then this shared task may have made a larger contribution to me than I made to it. It has however inspired a deeper interest in this problem so perhaps later work will have higher impact on the field itself.

As many of the articles appearing in the supplement are quick to acknowledge, suicide is a distressingly common cause of death particularly among younger people.  Has this work changed your view of suicide: do you find yourself more or less understanding or sympathetic of people who commit suicide and those they leave behind?  

I found the notes very moving. What was particularly tragic was the number of people who said they had to end their lives because they couldn't cope any longer with the physical pain they were dealing with due to various kinds of ailments. In some sense I felt like I understood why these people wanted to end their lives, and I could even believe it was the right decision, and yet it's hard not to feel a bit of despair in reading a note like that and imagining the circumstances under which it was written. I was also very moved by the simplicity of some of the notes - often rather brief with very simple instructions about paying a bill or describing where some important documents are located. I think it was the matter-of-factness about some of these notes that really struck me, and made me see suicide somewhat differently. Rather than being a very dramatic and emotional act, in quite a few cases it seemed like the writer of the note had simply reached the decision that this was a necessary thing for them to do. I think the simplicity of so many of these notes made them all the more human, and made me realize that people who commit suicide are perhaps not so different than anyone else. That was a sobering and important lesson for me.

View Dr Pedersen's personal homepage here

share on