Characterizing the psychiatric drug responses of Reddit users from a socialomics perspective

Social media has proven to be a safe space for people with mental illness to express themselves, a place where they are more willing to discuss their condition, treatment, and feelings. Thus, social media represents an important source of information for the analysis of the informal expression of the physical and mental responses to taking psychiatric drugs. In this paper, we propose a deep learning-based method to characterize drug reactions from a socialomics perspective. To this end, we construct seven base entity networks, one for each of five psychological entity types (affective, cognitive, perceptual, social, and personal concerns) and one each for side effects and disease. We then calculate the similarities between two entities (i.e., nodes) as the weight of the edges. Each node is represented by a combined vector consisting of semantic and graph embeddings. For each drug, we create a drug network and measure the variation in the network structure generated by adding the drug network to the seven base entity networks. If the variation in the network structure of a particular base network is larger than the others, it means that the drug has a larger impact on that base network. These results demonstrate that drug reactions can be assessed using social media, which may aid in the understanding of these reactions.