Depression detection in social media posts using affective and social norm features

Emotive language is also correlated with depression, as mental health issues affect the emotional state of people. It is empirically We propose a deep architecture for depression detection from established that depressed individuals express more negative social media posts. The proposed architecture builds upon thoughts, emotions and perspectives [9, 10, 11]. BERT to extract language representations from social media posts and combines these representations using an attentive Depression detection from social media can be performed bidirectional GRU network. We incorporate affective information, either at the individual post level or at the user level, given by augmenting the text representations with features extracted a collection of posts by said user. In [12], authors classify from a pretrained emotion classifier. Motivated by psychological depression-related LiveJournal posts, while in [5] authors focus literature we propose to incorporate profanity and on Twitter post classification. In [13], a shared task for CLPsych morality features of posts and words in our architecture using a 2015 is proposed for clinical diagnoses from Twitter posts.