Generative Conversational Agents- The State-of-the-Art and the Future of Intelligent Conversational Systems

Abstract

Intelligent conversational agents that generate responses from scratch are rapidly gaining in popularity. Sequence-to-sequence deep learning models are particularly well-suited for generating a textual response from a query. In this paper, I describe various generative models that are capable of having open-domain conversations. Toward the end, I present a null result I obtained in an attempt to train a chatbot from a small dataset and propose the use of a deep memory based machine translation model for training chatbots on small datasets.