Mental Helath-related dialogue data were taken from kaggle and apart from that non-mental health-related conversations for model effectiveness were added manually.
Non-mental health-related data were added so that the model can get confined to give response only related to mental health conversations.