Mental Health Analysis using Deep Learning of Social Media Data gathered using Chrome Extension

Recently the topic of ’mental health’ has gained high awareness, especially among younger generations. The age group that mostly gets affected by mental disorders is between 15-49 years of age. Research conducted on different social media data has found that a person with mental disorders (such as depression, anxiety, etc) tends to use social media more as compared to a mentally healthy person. Most users convey their emotional states through posting on public platforms, tweeting and accordingly watching related videos on YouTube that best describes their current mental state. It is also found that there is a difference in users’ social media activities (on Facebook, Twitter, and YouTube) with a healthy mind and one with an unhealthy mind. A study conducted in 2018 found that 45% of teenagers worldwide are constantly using the internet. According to research, teenagers use the internet on average for about nine hours each day. The internet can prove to be the easiest medium to monitor a person’s online activities (such as posting on social media, writing emails, watching or liking YouTube videos, commenting on YouTube videos, etc). The aforementioned activities and the internet usage time of a person can help to understand his/her mental state. The user’s social media activity can be analysed for finding potential patterns in his emotions/mood. These patterns can be helpful in uncovering underlying mental health issues such as depression, stress disorders, PTSD, schizophrenia etc. These findings can be reported to the user’s therapist for their monitoring record. To predict the mental state of the user by analysing their social media activities, an initiative was taken to develop an application with the objective to forewarn its users of any negative patterns regarding the state of mind that they exhibit. This application makes use of the Bidirectional LSTM algorithm, a deep learning method to identify a user’s emotions (one among the 6 defined emotions) and stores it for further usage. This application feeds the user’s data retrieved by the Chrome extension from 4 sources i.e. Twitter, YouTube, Gmail and Journal, to the trained model and stored the output. This output is displayed in a graphical format as well as in detail to the user. This application can become a personal companion to users by keeping track of their emotions and identifying any decline in their mental state and hence improving their emotional well-being.