Interaction and language patterns of an online depression communities in Korea
|
If you are the presenter of this abstract (or if you cite this abstract in a talk or on a poster), please show the QR code in your slide or poster (QR code contains this URL). |
Abstract
Background
Depression, which is a common and highly disabling disorder, causes high healthcare and societal costs. One key challenge in treating depression is at the early identification of individuals who suffer from depression, as these users tend to withdraw from real-life social activities. With the advent of social networking services, attempts have been made to detect early symptoms of depression on online space.
Objective
The purpose of this study was to characterize the interaction and language patterns of online communities that are dedicated to people with depressive moods. Understanding the unique characteristics of such communities can help identify other depressed users online for the ultimate goal of providing online healthcare interventions to those in need.
Methods
We obtained data from two popular online forums in Korea that are dedicated to subjects with depressive moods (depression forums) and other popular forums that focus on general topics such as foods and travel (general forums). Depression forum data include 3,634 posts by 1712 users and 20858 replies between 2006 and 2012, and general forum data include 2619 posts by 2619 users from the same period. For those who provided gender information in depression forums, 398 were males (mean age=29.7±6.1) and 616 were females (mean age=26.5±6.2). We examined the number of posts and replies per user and compared the language pattern in both forums with Naïve Bayes classification, a probabilistic algorithm widely used in pattern recognition.
Result
Several differences between depression and general forums emerged. First, most users in depression forums submitted their posts or replies at night (between 2000h and 2400h) with a peak at 2300h, while general forums were active throughout the day (1300h and 2400h). Second, similar to general forums, depression forums exhibited heavy-tailed behaviors in terms of the number of posts, where the majority of users (80%) contributed to at most two posts and a small fraction of heavy users (2%) contributed to more than 10 posts. The 10 most active commenters replied to over 100 posts. Third, depression forum users were getting emotional and informational support online. Every post received at least 3 replies, indicating an active support system within the community. Forth, the language used in depression forums clearly marked depressive moods, which differed from the language used in general forums. We observed an increase of negative sentiment, first-person pronoun, ellipsis, and emoticons related to sadness (like "TT"). The types of language posted online had high predictive power in identifying posts from depression forums from Bayes classifier achieving precision and recall of 0.85 and 0.98, respectively.
Conclusion
Online social activities are promising data for detecting early symptoms of depression. Given that a non-negligible fraction of users in depression forums talked about their past experience on clinical treatments, we believe online data can be utilized to extract linguistic markers of depression. Furthermore, heavy commenter can be adapted as moderators in depression forums, as people share their feelings and counsel each other without being conscious of others. In light of this finding, we would like to identify and handle depressive moods at large by developing software tools that automatically send out information about healthcare resources to interested online users whose language contains depressive markers.
Depression, which is a common and highly disabling disorder, causes high healthcare and societal costs. One key challenge in treating depression is at the early identification of individuals who suffer from depression, as these users tend to withdraw from real-life social activities. With the advent of social networking services, attempts have been made to detect early symptoms of depression on online space.
Objective
The purpose of this study was to characterize the interaction and language patterns of online communities that are dedicated to people with depressive moods. Understanding the unique characteristics of such communities can help identify other depressed users online for the ultimate goal of providing online healthcare interventions to those in need.
Methods
We obtained data from two popular online forums in Korea that are dedicated to subjects with depressive moods (depression forums) and other popular forums that focus on general topics such as foods and travel (general forums). Depression forum data include 3,634 posts by 1712 users and 20858 replies between 2006 and 2012, and general forum data include 2619 posts by 2619 users from the same period. For those who provided gender information in depression forums, 398 were males (mean age=29.7±6.1) and 616 were females (mean age=26.5±6.2). We examined the number of posts and replies per user and compared the language pattern in both forums with Naïve Bayes classification, a probabilistic algorithm widely used in pattern recognition.
Result
Several differences between depression and general forums emerged. First, most users in depression forums submitted their posts or replies at night (between 2000h and 2400h) with a peak at 2300h, while general forums were active throughout the day (1300h and 2400h). Second, similar to general forums, depression forums exhibited heavy-tailed behaviors in terms of the number of posts, where the majority of users (80%) contributed to at most two posts and a small fraction of heavy users (2%) contributed to more than 10 posts. The 10 most active commenters replied to over 100 posts. Third, depression forum users were getting emotional and informational support online. Every post received at least 3 replies, indicating an active support system within the community. Forth, the language used in depression forums clearly marked depressive moods, which differed from the language used in general forums. We observed an increase of negative sentiment, first-person pronoun, ellipsis, and emoticons related to sadness (like "TT"). The types of language posted online had high predictive power in identifying posts from depression forums from Bayes classifier achieving precision and recall of 0.85 and 0.98, respectively.
Conclusion
Online social activities are promising data for detecting early symptoms of depression. Given that a non-negligible fraction of users in depression forums talked about their past experience on clinical treatments, we believe online data can be utilized to extract linguistic markers of depression. Furthermore, heavy commenter can be adapted as moderators in depression forums, as people share their feelings and counsel each other without being conscious of others. In light of this finding, we would like to identify and handle depressive moods at large by developing software tools that automatically send out information about healthcare resources to interested online users whose language contains depressive markers.
Medicine 2.0® is happy to support and promote other conferences and workshops in this area. Contact us to produce, disseminate and promote your conference or workshop under this label and in this event series. In addition, we are always looking for hosts of future World Congresses. Medicine 2.0® is a registered trademark of JMIR Publications Inc., the leading academic ehealth publisher.
This work is licensed under a Creative Commons Attribution 3.0 License.