A hybrid system for online detection of emotional distress

Abstract

Nowadays, people are familiar with online communication and tend to express their deeper feelings on the Web. In the light of this situation, we present a hybrid system based on affect analysis for mining emotional distress tendencies from publicly available blogs to identify needy people in order to provide timely intervention and promote better public health. We describe the system architecture with a hand-crafted model at a fine level of detail. The model, which incorporates human judgment, enables the adjustment of prediction in machine learning on blog contents. The system blending supervised and unsupervised approaches will be examined and evaluated in lab experiments and practice. © 2012 Springer-Verlag.