Design And Implementation Of Intelligent Express Customer Service System Based On Task Oriented Dialogue

Intelligent voice conversation is one of the earliest branches of large-scale use in the field of artificial intelligence,it can realize multiple rounds of natural language conversations with human beings.Its core technologies have continuously developed and matured in recent years,including speech recognition,natural language understanding,and dialog management.At the same time,when the volume of the express logistics industry continues to increase and customer service encounters bottlenecks due to labor costs and other reasons,intelligent customer service system has become an excellent solution,which can not only quickly satisfy customers' needs,but also has 24/7 high-quality service and other advantages.Through in-depth investigation of the express delivery industry and the study of the theoretical technology about multiple rounds of dialogue,I designed and implemented an intelligent express customer service system that supports IVR navigation,express ordering,reminders checking,and information consultation.First the paper elaborates the background,value and current research status of the project at home and abroad,and then briefly introduces the key technologies such as ASR,NLU,DM,NLG,TTS and VUI design methods that will be used in the multiple rounds of dialogue system.Then,according to the requirement of the customer calling the customer service phone of the express company,the paper introduces the requirement analysis which establishes the functional model of the system.The paper builds the system architecture,divides the modules and explains the database and interface design in the outline design chapter.Then,the speech recognition model training method,natural language text classification algorithm model Text CNN,sequence labeling model LSTM+CRF,the dialogue process for specific modules of the system and interaction among systems are designed and implemented in detail.Then in the testing chapter,the paper records functional testing by automated scripts and SIP phone simulation,server stress testing by Jmeter and performance testing.Finally,the content of the paper is summarized and the outlook for future optimization is proposed.The system has served many express delivery companies,covering several cities,and currently has access to quite high user dialing traffic.In the actual operation process,the intelligent express customer service system shows high availability,and the resolution rate reached more than 70%,which truly achieved its business goals and value.