NEU-chatbot: Chatbot for admission of National Economics University

Abstract

In the last few years, intelligent chatbot systems have been prevalent in various application fields, especially in education. Therefore, the demand for such online consulting services like chatbots is getting higher respectively. However, most communications between potential students and universities are performed manually, which is very time-consuming procedure, becoming a burden on the head of admissions. In this paper, we introduce an AI-based chatbot where students can instantly get daily updates of curriculum, admission for new students, tuition fees, IELTS writing task II score, etc. Our chatbot was developed by Deep Learning models, which are already integrated into the Rasa framework. We also proposed a rational pipeline for Vietnamese chatbots with our data preprocessing to obtain optimal accuracy and to avoid the overfitting of the model. Our model can detect more than fifty types of questions from users' input with an accuracy of 97.1% on test set. The chatbot was applied for National Economics University's official admission Fanpage on the Facebook platform, which is the most famous social network in Vietnam. This research shows detailed guidelines on how to build an AI chatbot from scratch, and the techniques we used, which can be applied to any language globally.

Keywords

• Rasa
• Admission chatbot
• RNN
• BERT