Hindi Speech-Based Healthcare Chatbot

Abstract

Health care is critical to lead a healthy life. However, in cases of unhealthy conditions, consulting with the doctor is very difficult. The proposed idea creates a medical or healthcare chatbot using Machine Learning (ML) and Natural Language Processing (NLP). Speech recognition API and Google Translate API are used to recognize the user’s voice and text translation, respectively. Along with NLP, text classification algorithms facilitate fast, cost-effective and scalable solutions. ML algorithms are implemented on the corpus for prediction and classification. Text data classification is essential in order to minimize manual work and spend time reading the report. In this paper, a corpus of nearly 5000 documents is classified based on the symptoms and using five machine learning algorithms. The misclassification error is used in contrast with technical progress. Out of the six experimented algorithms, Gaussian Naïve Bayes classifier shows the highest accuracy score and out performs all others. The combination of sentence similarity and keyword matching algorithms are used to extract appropriate answers to the particular class and display it to the user. In general, users are not conscious of all the medications or the signs and symptoms of the actual condition. For minor issues, users have to take a time-consuming inspection directly at the hospital. A survey states that around 850 million of 1.3 billion people speak Hindi in India. The proposed chatbot is for these people who can only speak or understand the Hindi language. The speech-based healthcare chatbot algorithms are designed with query interpretation and user message comprehension. The machine learning algorithm is implemented to predict the class and with the combination of sentence similarity and keyword matching algorithms to provide the user with relevant answers.