A Smart Chatbot Architecture based NLP and Machine Learning for Health Care Assistance

Abstract

A chatbot or conversational agent is a software that can communicate with a human by using natural language. One of the essential tasks in artificial intelligence and natural language processing is the modeling of conversation. Since the beginning of artificial intelligence, its been the hardest challenge to create a good chatbot. Although chatbots can perform many tasks, the primary function they have to play is to understand the utterances of humans and to respond to them appropriately. In the past, simple statistic methods or handwritten templates and rules were used for the constructions of chatbot architectures. With the increasing learning capabilities, end-to-end neural networks have taken the place of these models in around 2015. Especially now, the encoder-decoder recurrent model is dominant in the modeling of conversations. This architecture is taken from the neural machine translation domain, and it performed very well there. Until now, plenty of features and variations are introduced that have remarkably enhanced the conversational capabilities of chatbots. In this paper, we performed a detailed survey on recent literature. We examined many publications from the last five years, which are related to chatbots. Then we presented different related works to our subject, and the AI concepts needed to build an intelligent conversational agent based on deep learning models Finally, we presented a functional architecture that we propose to build an intelligent chatbot for health care assistance.