A Comprehensive Approach To Healthcare Chatbots Using SpaCy NLP And Naive Bayes Algorithm

Imagine a world where medical assistance is just a conversation away, where artificial intelligence helps navigate your health concerns with precision, by the intersection of machine learning and healthcare, our chatbot emerges as a powerful ally for patients everywhere. This research work focuses on developing a healthcare chatbot which provides diagnosis of patients by taking user-input symptoms, suggesting medications and also hospitals in a particular region. Both speech and text options are introduced to interact with the user. Firstly, the user is prompted to enter the symptoms faced by them, the symptoms are extracted from user input sentences using SpaCy NLP. These extracted symptoms are fed to a Multinomial Naïve Bayes Classifier for disease prediction. Additionally, chatbot also provides the required medications and the hospitals in the region from the data stored in the CSV and PDF files. The user interface of the chatbot is built using Flask. SQL is leveraged for user account management and to store the diagnosis history of the user. The dataset used for training is a CSV file consisting of 132 unique symptoms, 42 unique diseases and a total of 4,920 combinations of symptoms and associated diseases. The dataset is pre-processed in the form of 1s and 0s to make it suitable for our Multinomial Naive Bayes model which achieved a test accuracy of 95% which is better than the existing models having an accuracy of around 93%.