Artificial Intelligence Chatbots: A Survey of Classical versus Deep Machine Learning Techniques
Date
2023
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
: Artificial Intelligence (AI) enables machines to be intelligent, most importantly using Machine Learning (ML) in which
machines are trained to be able to make better decisions and predictions. In particular, ML-based chatbot systems have been developed
to simulate chats with people using Natural Language Processing (NLP) techniques. The adoption of chatbots has increased rapidly in
many sectors, including, Education, Health Care, Cultural Heritage, Supporting Systems and Marketing, and Entertainment. Chatbots
have the potential to improve human interaction with machines, and NLP helps them understand human language more clearly and thus
create proper and intelligent responses. In addition to classical ML techniques, Deep Learning (DL) has attracted many researchers to
develop chatbots using more sophisticated and accurate techniques. However, research has paid chatbots have widely been developed
for English, there is relatively less research on Arabic, which is mainly due to its complexity and lack of proper corpora compared
to English. Though there have been several survey studies that reviewed the state-of-the-art of chatbot systems, these studies (a) did
not give a comprehensive overview of how different the techniques used for Arabic chatbots in comparison with English chatbots; and
(b) paid little attention to the application of ANN for developing chatbots. Therefore, in this paper, we conduct a literature survey of
chatbot studies to highlight differences between (1) classical and deep ML techniques for chatbots; and (2) techniques employed for
Arabic chatbots versus those for other languages. To this end, we propose various comparison criteria of the techniques, extract data
from collected studies accordingly, and provide insights on the progress of chatbot development for Arabic and what still needs to be
done in the future.