RAAH.ai: An interactive chatbot for stress relief using Deep Learning and Natural Language Processing
Romit Vinod Kankaria, Aman Agrawal, Harshit Barot, Anand Godbole · 2021
the ongoing lockdown where the accessibility to people you want to interact with or who could provide to you the much needed relief has drastically reduced, be it doctors or therapists. Keeping this in mind, along with the huge amounts being spent by people on consultation, we thought of coming up with a system which would be relevant to people going through a tough time. There is always a need for people to talk with their loved ones even when they are not around and our voice cloning system would serve the purpose. There is a need for real-time artificial human conversation that is customized and personalized according to a person's needs. In addition, there is a requirement of security for preventing voice spoofing which might lead to malpractices. This led us to materialise a solution which would allow a person to interact with people. The system uses NLP methods of chatbot to identify the emotion/problem that the user is trying to express. It uses CNN and RNN to respond to the user in the voice of a loved one. Using the LibriSpeech and VCTK datasets, a Tacotron-based architecture was generated by us which resulted in a similarity percentage of 91% along with a response time of 4 seconds which came out to be better than the various methods studied during the literature survey.