Abstract:

Conversational systems such as chatbots or virtual assistants play an integral role in our life. The usage of chatbots has expanded significantly in industry and education. With the advancement of technology in artificial intelligence, chatbots have been developed from a rule-based approach to a neural network approach to have a human conversation. A virtual assistant as a tutoring agent can mimic humans and provide the best replacement for a human tutor. It makes users search easier by answering the frequently asked questions, by providing suggestions to improve the task performance interactively and effectively in a short time. Most universities have started deploying chatbots as learning assistants to help students with queries in less time and are available 24/7. It reduces the workload of the lecturer along with retaining the synchronous communication between the students and tutor. Recently conversational AI systems have achieved great progress on various tasks, such as Intent classification, entity extraction, and sentiment analytical analysis with the use of cutting-edge natural language processing and neural network processing technology. RASA is an open-source NLU and DIET model implementation. The tutoring agent as a chatbot designed in this research provides an accurate answer for any query based on the dataset of Frequently Asked Questions. The provided dataset is related to a research seminar course available for the students of Computer Engineering at the Chemnitz University of Technology. The developed chatbot model can assist students in answering all the questions, from seminar course enrollment through its method of completion. The chatbot replies to the student query by using natural language processing. The chatbot is deployed using a Python framework called RASA, it has built-in libraries for natural language understanding and dialogue management. The bot learns from the environment to predict the close response. The results represent the bot's accuracy in user intent prediction and interactive response generation. These are demonstrated from the confusion matrix and histogram of the trained model.

Keywords:

Chatbot, Tutoring AI agent, RASA, Natural Language Processing, Dialogue Management, Natural Language Understanding.