基于自适应多叉树防碰撞算法的智能客服NLP短文本分类模型  

Intelligent Customer Service NLP Short Text Classification Model Based on Adaptive Multi-tree Anti-collision Algorithm

作　　者：李俊峰 黄秀彬 刘娟 董蓓 王建波 LI Junfeng;HUANG Xiubin;LIU Juan;DONG Bei;WANG Jianbo(China marketing department,State Grid Corporation,Beijing 100000,China;Customer Service Center of State Grid Co.,Ltd.,Tianjin 300304,China;Beijing China-Power Information Technology Co.,Ltd.,State Grid Information&Telcommunication Group,Beijing 100085,China)

机构地区：[1]国家电网有限公司,市场营销部,北京100000 [2]国家电网有限公司客户服务中心,天津300304 [3]国网信通产业集团北京中电普华信息技术有限公司,北京100085

出　　处：《微型电脑应用》2023年第1期45-48,共4页Microcomputer Applications

基　　金：国网客服中心2020年知识库省公司应用功能拓展项目(659931190051)。

摘　　要：智能客服的工作方式是读取客户发送的短文本信息,对客户的问题进行解答。为了提高智能客服对于短文本的理解能力,提高智能客服的服务质量,提出基于自适应多叉树防碰撞算法的智能客服NLP短文本分类模型。根据短文本分类模型的工作任务,重新设计智能客服NLP短文本分类模型的核心架构,逐步分析自适应多叉树防碰撞算法的工作原理和内容、短文本分类模型的网络通信方式,调用融合卷积神经网络、多头自注意机制以及以上提到的技术,集成基于自适应多叉树防碰撞算法的智能客服NLP短文本分类模型。实验结果表明,该模型的处理短文本分类结果的准确率、精确率、召回率皆符合预期设计的目标。The working mode of intelligent customer service is to read the short text messages sent by customers and answer their questions. In order to improve the understanding ability of intelligent customer service for the short texts and improve the service quality of intelligent customer service, an NLP short text classification model of intelligent customer service based on adaptive multi tree anti-collision algorithm is proposed. According to the task of short text classification model, we redesign the core structure of the intelligent customer service NLP short text classification model, analyze the working principle and content of the adaptive multi tree anti-collision algorithm and the network communication mode of the short text categorization model. We use the fusion volume integral neural network, the multi head self attention mechanism and the above mentioned technology. An intelligent customer service NLP short text classification model based on adaptive multi tree anti-collision algorithm is integrated. The experimental results show that the accuracy and recall rate of the short text classification are in line with the expected design goals.

关 键 词：自适应多叉树防碰撞算法 短文本分类模型 分类模型架构 网络通信方式 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]