Asian Journal of Pure and Applied Mathematics

There is no gain-saying the fact that the AI technology is the spur of the moment with high prospects into the future. The sub-field of Deep Learning in AI is making giant strides in all sectors of life. Inclusively, conversational AI-based Chabot is not an exception in this milestone. The techniques behind an AI Chabot is to understand patterns and add context to a conversation in a unique concept so as to expand the boundary of knowledge. The emergence of deep learning models as the state-of-art of all models - ranging from Neural models to the Pre-trained Language Models (Transformer, BERT and GPT) is not a surprise due to its capability to learn from large textual datasets and fine-tuning. This article reviews the model of extraction of features and word embedding techniques that could yield a perfect context retention. Although many traditional or rule-based Chabot are so common, a clue to understanding the models of Natural Language Understanding (NLU) and Natural Language Generation (NLG) yields an exploration into Deep Learning-based contextual AI-based Chabot. Consequently, the mechanism for input sequences enables the model to capture contextual information, long-range dependences, learning the inherent structure and pattern of user’s input to yield a contextual response. Decision making in business can be efficiently reached if intelligent systems can achieve an improved, personalized and scalable customer service due to context retention.