The Evolution and Frontier Exploration of Chatbots Driven by Natural Language Technology

As an intelligent system taking natural languages as its medium of interaction, the development of chatbots vividly reflects the development process of natural language technology. From those earliest programs that could only carry out simple question-and-answer procedures to the modern intelligent assistants that can engage in multiple turn complex conversations, each advance of chatbots has closely correlated with the transformations in natural language understanding and creation technologies. This paper divides it into 4 phases according to the time order, which are the rule-driven phase, the statistics learning phase, the deep learning phase, and the large language model phase. Detailed explanation for major technical characteristics, typical applications and technical restrictions of every stage. With some specific experiments and applications, it shows how chatbots start to overcome problems such as unclear languages and context dependencies. Finally, based on the trends of technology, this paper makes a guess of where the current chatbots are heading towards general intelligent interaction, and it attempts to explore those challenges today’s technologies and tomorrow’s technologies may have to meet and it is intended to provide some reference to help more people know about the technology in this aspect.