A NATURAL LANGUAGE UNDERSTANDING KNOWLEDGE-BASED CHATBOT OVER LINKED WEB DATA

With the abrupt development of the web today, the information in web pages is structured and tagged to ensure that it can be read directly by computers, which is available publicly in the type of knowledge bases(KBs). Ensuring this information is machine-readable, easily reachable, and valuable for a customer is one of the goals of the chatbot over linked web data in real-time. Developing a knowledge-based chatbot over linked data leads to many challenges, including user question understanding, multilingual recognition, and supporting multiple knowledge bases. To deal with these challenges, we propose a solution; 1st we design and develop interactive user interface system architecture, 2nd, we plan a machine learning(ML) method for natural language understanding(NLU) to be familiar with user queries via text/voice message methods, then take out the necessary keywords. Information for those keywords is obtained by making queries to the linked website's knowledge base to get desired information using the web scraping technique. For the experiment, we use generic medicine data as a domain, and it's added to the frequently asked queries dataset in the system knowledge base, so the existing system knowledge base gets extended automatically by the chatbot intelligent features, such as addressing the user’s queries, feedback messages, and the system failed responses. This system can be widespread with the new domains for an extensive range of topics