摘要
随着电力用户规模不断增长、业务量不断扩大,传统客服模式已不能满足繁重的用户业务咨询需求,在此背景下设计一种精准的客户服务模式就显得至关重要。针对这一需求,文章首先提出了一种融合注意力机制+双向长短期记忆网络+双向门控循环单元(Attention+BiLSTM+BiGRU)的序列到序列(Seq2Seq)模型;然后通过对上下文语义信息的深层时序特征提取及赋权,有效提高了模型对话性能;最后通过算例仿真,实验结果从定性和定量的角度验证了所述方法具有回答效果更佳、问题识别能力更高的特点,对在线电力智能客服系统的设计和实现具有一定的参考价值。
With the growing scale of power users and the continuous expansion of business volume,the traditional customer service model can no longer meet the heavy user business consulting needs.To solve the problem,this paper proposes a Seq2Seq model that integrates Attention+BiLSTM+BiGRU.By extracting and weighting the deep temporal features of context semantic information,the model dialogue performance is greatly improved.The experimental results verify that the PROPOSED MODEL has better answer effect and higher problem identification ability from the qualitative and quantitative perspectives,which has a certain reference value for the design and implementation of online power smart customer service system.
作者
李刚
张曦月
杨维
LI Gang;ZHANG Xiyue;YANG Wei(Department of Computer,North China Electric Power University,Baoding 071003,Hebei Province,China;Engineering Research Center of Intelligent Computing for Complex Energy Systems,Ministry of Education,Baoding 071003,Hebei Province,China;Customer Service Center,State Grid Corporation of China,Dongli District,Tianjin 300309,China)
出处
《电力信息与通信技术》
2023年第8期68-74,共7页
Electric Power Information and Communication Technology
基金
国网客服中心–电力营销业务–2020年网上国网服务后台–设计开发实施项目(71993118000D)。
