Knowledge Graph Based Medical Chatbot building

To have a good and unproblematic life without any health risks, it is very important to get medical advice on any health-related problems. However, getting medical advice incurs costs. Chatbots are AI/ML based software which may be trained with a lot of inquiries and responses and match users’ inquiries against a large repository of evidence-based medical data to provide simple answers. To reduce the healthcare costs and improve accessibility of medical knowledge a medical chatbot can be built using ML and NLP techniques. In the existing system of such chatbots several databases are connected together using join statements making it more complicated to access the data. The medical knowledge is vast and varied hence giving many disadvantages to use fixed schema. In this paper we have proposed a knowledge graph based method of chatbot creation for the healthcare field. When the patient enters the symptoms, they are suffering from, then chatbot evaluates and based on the evaluation recognizes the disease. The basic idea of our work is to build a chatbot which can evaluate the symptoms and using this evaluation, rank the possible disease which the patient could be suffering from. The chatbot gets its knowledge from a knowledge graph built on an extensive TigerGraph database. The data from the TigerGraph database can be accessed using different analysis queries. The chatbot will be considered profitable only when it can diagnose all kinds of disease and provides the necessary advice measures to be taken. The knowledge graph used here can be improvised by including results from lab tests to arrive at a better diagnosis and increase the precautions by including the medicines to be taken. The chatbot can only be as smart as the knowledge graph database and hence the evaluations made must be checked with the medical professional. The chatbot has good accuracy in predicting the disease the user is suffering from.