Deep Learning for Dialogue Systems: Chit-Chat and Beyond

Starting from the 1960s, conversational artificial intelligence has become a crucial research field and has grabbed much more attention in recent years. Empowered by deep neural models, dialogue systems have demonstrated very impressive and appealing performance in virtual assistants and social bots. In viewing its potential and values, mainstream NLP, IR, and even ML communities have started contributing to dialogue systems. Dialogue systems can be roughly grouped into two classes, i.e., task-oriented and chit-chat systems. The former group focuses on completing predefined tasks with task-specific constraints and goals, e.g., restaurant booking and making calls. The later systems are mainly designed...