A Comparative Study of ChatGPT and Seq2Seq Chatbot for Effective Students Advising

Abstract

AI chatbots are increasingly becoming a part of students’ lives. From school learning to universities admission and orientations. AI chatbots are finding ways to facilitate and improve learning experiences. Moreover, AI chatbots employ state-of-the-art deep learning algorithms, with natural language processing (NLP). For example, Seq2Seq based chatbots adopted the deep learning algorithms powered by recurrent neural networks (RNNs). Moreover, BERT model is the first language model that adopted self-attention mechanism by using the bidirectional Transformers. the GPT (Generative Pre-Trained Transformer) is another advanced language model that trained by using the autoregressive model. This paper aims to focus on domain-specific educational purposes and particularly for advising prospective undergraduate students, by answering to their enquiries such as universities admission tests, applications deadline and selecting the best fit majors and universities. Additionally, this paper compared the architecture and performance in both the ChatGPT and Seq2Seq chatbot models for providing advices to prospective students.