基于Seq2Seq的智能客服机器人的构建

Seq2Seq Based Intelligent E-commerce Customer Service Chatbot Construction

针对目前淘宝、京东等电商店铺客服依旧是以人工服务为主,基于规则匹配的回复为辅的现状,提出基于Seq2Seq模型实现电商店铺客服的自动化回复系统,降低人工成本,也尽可能为用户带来实时性的客服体验。在实现时基于三层模型处理,分别通过基于正则表达式的任务式回复、基于特征库匹配的检索式回复以及基于Seq2Seq的生成式回复为客户提供精细良好的服务。最后配备上服务器来处理问答需求,运行模型,建立基于安卓UI的回复界面以及敏感词转人工客服机制,减少人工客服的处理量,为电商店铺及顾客带来更好的客服体验。

Aiming at the current situation that the customer service of e-commerce stores such as Taobao and JD is still based on manual services,and the response based on rule matching is supplemented,it is proposed to use the Seq2Seq model to implement an automated response system for e-commerce store customer service,reduce labor costs,and do the best possible Users bring real-time customer service experience.Three layers of processing are used in implementation,and more refined and better services are achieved through task-based responses based on regular expressions,search-based responses based on feature library matching,and generative responses based on Seq2Seq.Finally,it is equipped with a server to handle Q&A needs,run models,establish Android UI response interfaces and a mechanism for transferring sensitive words to manual customer service,reducing the amount of manual customer service processing,and bringing a better customer service experience to e-commerce stores and customers.