Interactive Chatbot for Improving the Text Classification Data Quality

The pandemic is affecting the global community in many ways. In most developing countries, there is a limitation in the detection facilities, which affect many suspected cases. This paper proposes a chatbot framework to assist and provide guide for the suspected/infected patients with COVID-19. Conversational software agents activated by natural language processing is known as chatbot, are an excellent example of such machine. Our COVID Bot is based on an integrated model between the rule-based model and the class classification model, having the rule-based model integrated with the MongoDB NoSQL database. Chatbot, using Natural Language Processing (NLP) and data mining techniques to assist patients by providing immediate answers for their questions. It also acts as a novel communication mean for impaired people for sharing knowledge and information, through conversing with them. Based on the literature review, this paper compared our methods with three classical classification algorithms: random forest, gradient boosting, and multi-layer perceptron (MLP). Experimental results show that our proposed chatbot greatly improves the classification performance, with IE-Net as 94.80%, 92.79% as recall, 92.97% as precision and 94.93% as AUC for distinguishing COVID-19 cases from non-COVID-19 cases (with only common clinical diagnose data).