随着客户规模的不断增长，服务量也逐年水涨船高，人工服务的成本劣势逐渐暴露
出来。正是在这种背景下，在线客服逐渐被作为一个降低服务成本的有效手段被采
用。当前常见的在线客服模式主要有三种：基于第三方系统平台加载公司内部知识
库的模式、基于公司官网加载内部知识库的模式、基于社交平台的互动服务模式。

但实际上，无论是那种模式，都难以保证服务质量能够与热线服务相媲美，无法起
到替代热线人工服务的作用。

广东移动在线客服的建设思路是建设一个高效低成本的在线服务模式，同时能
够确保客户的服务感知不下降。本文重点阐述了广东移动在线客服的模式。该模式
基于 DSS 模型——以自然语言识别技术分解客户的文本问题并加以识别和理解，同
时建设一套客户化的知识库系统作为模型库，以公司内部系统和客户问题中蕴含的
信息作为数据库的信息来源，采用专业的索引和搜索技术作为方法库。在实际服务
过程中，通过自然语言识别技术理解客户的问题后，结合数据库提供的相关信息，
以搜索的方式获取模型库中的答案模板并推送给客户，从而完成服务。最后，本文
对广东移动在线客服的应用情况进行了分析，进一步论证了广东移动在线客服模式
的成功。

关键词：在线客服；自然语言识别；知识库；决策支持系统
Abstract
It has been more than ten years since call center entered China. The traditional
hotline services occupy the mainstream position. With the growing scale of customers,
the service volume goes up year by year, and the disadvantage of cost on manual service
has been exposed already. Under this background, online services have been adopted as
an effective means to reduce the cost of service. Actually, there are three main patterns of
online service: the pattern loading the internal knowledge base on a third-party platform,
the pattern loading the internal knowledge base on the company’s official website and the
pattern based on social mediea. In practice, however, whatever patterns to be adopted, it
is difficult to guarantee that the quality of service can be as good as the hotline service. It
can’t replace the hotline service.
It is the guiding ideology of building the online service of China Mobile Group
Guangdong Corporation(GMCC for short) to build an efficient and low cost online
service, which will not drop customers’ aesthesis. This article focuses on the online
service pattern of GMCC. This pattern is based on DSS, recognizing and understanding
the customers’ text question using Natural Language Processing, building a customized
knowledge base as the model base, using the company’s internal systems and the
information in the customer text questions as the information source of database and
adopting professional indexing and searching technology as the method base. In practice,
after understanding the customers’ questions with NLP, combining the related
information of database, it will search for the answer template in the model base, and
send it to customers to complete the service. At last, we analyze the application situation
of online service pattern in GMCC, which will demonstrate the success of the pattern.
Key Words：Online Service; Natural Language Processing; Knowledge Base; Decision
Support System