Using Deep Linguistic features to predict Depression
Thang Nguyen, William W. Armstrong, Anilesh Shrivastava · 2013
We participated in an in class shared task of predicting CES-D Depression Score by using linguistic features on Facebook Statuses of the users. Our attempt was to look beyond surface features like bag of words and Topic model. To that end we explored several off the shelf tools for parts of speech tagging, Supervised LDA, deep learning and active learning. In this paper we document our experience of using such tools, and challenges we faced while building this prediction system. In the end we were able to achieve good results through supervised LDA and got some promising insight into deep learning.