Based On The Seq2seq Reservation System Intelligent Response Research And Application

With the rapid development of big data and artificial intelligence technology and the continuous deepening of "Internet +" actions,business intelligence has risen rapidly in all walks of life.In order to reduce the operating costs of enterprises,in the field of customer service,the traditional manual customer service mode has gradually changed to the electronic customer service mode.The intelligent customer service represented by chat robot has become the first choice for enterprise business intelligence assistants.In this paper,the research on intelligent response is based on the reservation system of Shuidong Township,a precision poverty alleviation project in Guizhou Province,and the research on the intelligent customer service application target of the reservation system proposed by the initial project needs.The traditional customer service system for search needs to maintain a large static knowledge base,which is easy to generate knowledge blind spots and affect the quality of service.Under the pressure of increasing system users,dialogue-generated intelligent customer service based on natural semantic understanding comes into being.Since the service customer process will generate a lot of new knowledge due to the customer's demand for consulting expression,the intelligent customer service can learn more self-training and generate more accurate answers according to different scenarios,which undoubtedly improves the application value of the project.Based on the Seq2 Seq dialog generation model,this paper designs and implements the “text + operation guidance” response type reservation system intelligent response function by using the long-short-term memory network(LSTM)and the tensorflow deep learning framework in the cyclic neural network.The author first builds a Q&A corpus by accumulating and collecting relevant data in the field of B&B through the pre-run of the system.Secondly,based on the comparison of Seq2 Seq,Seq2Seq+Attention,Seq2Seq+Attention+Beam Search,the evaluation model based on BLEU,Perplexity and Distinct language models.The index evaluation model determines the superiority of the Seq2Seq+Attention+Beam Search model in dialogue generation.Finally,it proposes an effective solution for the problem of hyperparameter setting,gradient dissipation and gradient explosion in deep learning training,and realizes the intelligent system of “Shuidong Xiangshe” reservation system.Answer function.The intelligent response function realized in this paper can basically meet the expected goal of the intelligent customer service of the reservation system,and will improve the status quo of recruiting a large number of manual customer service in the reservation service,bringing more benefits to the enterprise.