Mental Health Chatbot The dataset contains Tags, Patterns (Questions asked to the chatbot) and Responses (Corresponding Solutions). The patterns are lemmatized so that questions with similar lemma can be correctly answered. The model is trained with the help of Neural Networks (Tensorflow, Keras) which helps us predict Responses for related Patterns. Sentiment Analysis is done with the help of transformers (BERT). If the model classifies tha chat to be high risk, it is directly connected to the nearest distress helpline. If the model classifies tha chat to be moderate risk, the case is matched to the right therapist. If the model classifies tha chat to be low risk, digital solutions are provided to the user.
