Abstract

With the ever-increasing usage of social media for either explicitly seeking help or simply sharing thoughts and feelings, we, in the computational disciplines, have the opportunity to utilize such data for building datasets, models, and doing analysis.  In this talk, I will share our collaborative work done at the Information Retrieval Lab at Georgetown University with our ex/current doctoral students and outside collaborators on two similar yet different type of social media for mental health.

The first part of the talk focuses on the dedicated mental health forums where the users who register to share and communicate their thoughts and feelings are suffering from some sort of mental distress (sadness, depression, potential of self-harm….). The task is to triage the severity of users’ posts to be able to detect early the potential of self-harm. This effort was as the results of several years challenge, set up by the Computational Linguistics and Clinical Psychology (CLPsych), using data provided from a given mental health forum. We will also provide analyses of data to show the impact of forum activities and conversations on the users during period of time.

The second part of the talk focuses on the question of whether we can detect if a user is suffering from any one or more of the nine mental health conditions,  only using the *general language* of the user; that is, the posts are not in mental health [sub]forums nor have any mental health related words. For addressing this question we had to construct large scale datasets (RSDD: self-diagnosed with depression: 9K, control: 107K;  SMHD (self-diagnosed; 9 conditions): 37K, control: 336K). A subset of data is annotated for the temporal analysis of the conditions (RSDD-Time).  We have made our data available via Data Usage Agreement.   I will explain how we have identified the diagnosed users, and how selected carefully the controls.  Further I will show the results of several baselines to detect the conditions.

I will conclude with glimpses of additional health related projects of our lab.

Biography

Nazli Goharian is Clinical Professor of Computer Science at Georgetown University, and Associate Director of the Information Retrieval Lab . Prior to joining Georgetown (2010), she was Clinical Associate Professor of Computer Science and a member of the Information Retrieval Laboratory at the Illinois Institute of Technology (IIT). She joined academia from industry. Her research, publications, and doctoral student mentorship span the domains of information retrieval, text mining, and natural language processing. Specifically, her interest lies on humane-computing applications; as such, she has been focusing on text processing in medical/health domain. She has over 60 refereed conference and journal publications; she publishes in both the ACM and ACL communities. Joint with her doctoral students, she received a EMNLP 2017 Best Long Paper Award.  For her contributions to undergraduate and graduate curriculum development and excellence in teaching she was recognized with the College of Science and Letters Dean’s Excellence Award in Teaching in 2005, and in 2002, 2003, and 2007, with the Computer Science Department Teacher of the Year Award. In 2009, she was awarded the IIT Julia Beveridge Award for faculty (female faculty of the year).