Research On Intelligent Dialogue Policy For Airline Customer Service

With the rapid development of natural language processing techniques, people put forward higher requirements for Question Answering System(QA System). People are no longer satisfied with a simple QA system, and require advanced intelligent dialogue processing systems. Under the rapid development of e-commerce, the demand for network customer service is rising, and automatic QA Systems for e-commerce network customer service have emerged. Using intelligent automatic dialogue network customer service to supersede manual network customer service is the future development trend. Automatic dialogue network customer service not only saves labor costs for e-commerce industry, but also makes the network customer services more efficient.The focus of this paper is the intelligent dialogue policy selection for airline customer service. In order to engage in the intelligent dialogue policy for airline customer service, studies are also conducted on the dialogue system maintenance method and the design of robustness for a dialogue system. Dialogue policy is to guide dialogue process to carry out with a certain program, and an effective dialogue policy will agree on the same dialogue goals with the user quickly and guide the user to obtain a query need quickly. The complete interaction process is constituted by a series of questions and answers, so a network customer service dialogue system needs to maintain a complete dialogue process with dialogue management methods. A good design for robustness is also very important for a dialogue system, a robust system can deal with a wider variety of input descriptions and anomalies.This paper proposes two dialogue policies: the dialogue policy based on n-gram model and one based on information attributes constraints. The dialogue policy based on the n-gram model simulates the real dialogue process based on the real dialogue corpus which can model the real dialogue process. The dialogue policy based on information attributes constraints directs the dialogue according to the results of query feedbacks information entropy. This paper combines the two policies and builds a dialogue system based on the hybrid policies. To some extent, the dialogue policy can efficiently achieve the information query goals by guiding the dialogue process.