Expert System-based Chatbot as a Virtual Consultant






Most of the conversational chatbots were developed using Natural Language Processing and Machine Learning method, in which it tries to understand human natural language than trying to guess the response according to its trained knowledge-based. The drawback of this method is that in fact, this kind of chatbot will act as a human customer service without any technical deep knowledge for a specific area like technical support. Moreover, it will take time to train the chatbot, and approximately would be the same as the manual knowledge acquisition on an expert system. And, if it needs to add more knowledge, it needs to retrain again, while on an expert system, only input more data and will be ready immediately. Furthermore, the conversational chatbot will have difficulties understanding non-formal language by the user.

The non-conversational expert system based-chatbot would have a different characteristic when interacting with the user. On the one hand, the conversational chatbot will receive input from users and try to understand and respond to user questions. On the other hand, on the expert system-based chatbot, the user would only need to enter keywords or questions once, then the chatbot would be more active by asking the yes-no questions to the users. Moreover, the non-conversational chatbot could act as an expert rather than just as a customer service agent.

This research presents a different approach to developing a chatbot by using an expert system approach. The chatbot is also flexible to handle a different problem area depending on the entered knowledge-base subject. A framework was also developed to be able to produce a generic expert system-based chatbot. Moreover, to enhance more features, the user has options to select the inference mechanism from backward chaining, forward chaining and Dempster-Shafer method.



Expert System, Chatbot, Backward Chaining, Forward Chaining, Dempster-Shafer, Framework.