Multilingual Healthcare Chatbot Using Machine Learning
The proposed solution describes a multilingual healthcare chatbot application that can perform disease diagnosis based on user symptoms and responds to user queries by calculating sentence similarity by using TF-IDF and Cosine Similarity techniques and choosing the most appropriate response from its knowledge database.
Abstract
The healthcare sector is one of the largest focus areas in the world today. Health problems are becoming increasingly common. India faces a huge challenge in terms of managing rural healthcare. Early diagnosis and treatment of diseases can play an instrumental role. Physical consultation, particularly in the rural areas, is costly and time consuming. The solution is adopting healthcare chatbots. The proposed solution describes a multilingual healthcare chatbot application that can perform disease diagnosis based on user symptoms. It also responds to user queries by calculating sentence similarity by using TF-IDF and Cosine Similarity techniques and choosing the most appropriate response from its knowledge database. The multilingual capabilities of the chatbot system make it highly suitable for use in rural India. The chatbot application currently supports three languages namely English, Hindi and Gujarati. The chatbot application converses with the user using concepts of Natural Language Processing and also supports speech to text and text to speech conversion so that the user can also communicate using voice. Five different Machine Learning algorithms have been analyzed for disease prediction. The Random Forest Classifier produces the best results and gives an accuracy of 98.43%. Thus, it is used as the system’s core classifier.