Review of Deep Learning approaches for Conversational Artificial Intelligence

Conversational Artificial Intelligence (CAI) is now part of our life as we can find conversational agents everywhere in every kind of devices which make our life easy. Conversation assistants like Amazon Alexa, I-Phone Siri are now readily available. But development of these agents is not that easy, to model an agent that can understand the input (text, voice, and image) in natural language is a difficult task for a machine. We are in this field since long, but this area still remains the hardest one to perform. In our paper we have given a review about how this technology has grown till date. From handwritten rules to train the machine through deep learning methodology CAI has grown beautifully but has yet to blossom. The reason behind the slow growth of CAI is the natural language of humans, every country, states and even cities inside the state have different language or the different accent moreover to mimic the trait of humans in machines to naturally think about any topic in this world and to respond to same is the hardest part. In the end we have discussed how transformer-based model has improved the accuracy and future work regarding the same.