AN APPROACH FOR BUILDING A CHATBOT SYSTEM FOR THE ADMISSION PROCESS OF DA LAT UNIVERSITY

Abstract

A chatbot is a computer program designed for the chat interaction between robots and humans automatically using natural language techniques. This program is developed to be a virtual assistant which lures the human into the thought that they are talking to a real person. In this paper, we develop a chatbot system for the admission process of Da Lat university that allows the staff to answer questions immediately and automatically from users anytime. The important feature of any chatbot is to understand the user’s questions and to respond with appropriate answers. Our approach builds a chatbot application that adapts to the university's needs. We apply some BERT-based representation language models to predict the answer from the input question. The experimental results show that the salti/bert-base-multilingual-cased-finetuned-squad is a suitable model for our chatbot application since its F1 and EM scores for dev_set are remarkably high, accounting for 88.6% and 79.6%, respectively. For the intent classification, we achieve the accuracies of 99.9% and 100% for the validate accuracy and the test accuracy.