Cognitive distortion based explainable depression detection and analysis technologies for the adolescent internet users on social media

Nowadays, adolescents would like to share their daily lives via social media platforms, which presents an excellent opportunity for us to leverage these data to develop techniques to measure their mental health status, such as depression. Previous researches focus on the more accurate detection of depression through statistical learning and ignore psychological understanding of depression. However, psychologists have given lots of theoretical evidence for depression. Such as according to cognitive psychology research, cognitive distortions will result in depression. Thus, in this study, we propose a new task, explainable depression detection, to not only automatically detect depression but also try to give clues to depression based on cognitive distortion theory. For this purpose, we construct a multi-task learning model based on a pre-trained model to detect depression and identify cognitive distortion. And we use many analytical means including word clouds for data analysis to draw our conclusion. Previous social media users' depression corpus and our cognitive distortion corpus are utilized for analysis and experiment. Our experimental results outperform the baseline results and interesting conclusions about adolescent depression are drawn.