摘要
为提升电力行业客服系统的智能化水平和客户满意度,文中设计了一种基于卷积神经网络(CNN)和注意力机制(AM)的电力智能客服系统。该系统通过语音满意度识别算法,能够准确处理和理解电力行业特定的语音对话数据,从而实现精准的客户需求识别和高效的服务响应。在模型设计中,提出了一种融合多层CNN组件和注意力机制的语音特征处理方法,将输入语音数据的Mel频率倒谱系数、波形特征和语音频谱图等多种特征进行并行编码,并通过注意力机制分析这些特征的重要性。实验结果表明,所提出的语音满意度识别模型的性能显著优于传统的深度学习算法,准确率可达98.31%,比表现最佳的BiLSTM高5.58%。
In order to improve the intelligence level and customer satisfaction of the power industry customer service system,a CNN and AM based power intelligent customer service system is designed in this paper.The system can accurately process and understand specific language data in the power industry through speech satisfaction recognition algorithms,thereby achieving accurate customer demand recognition and efficient service response.In the model design,a speech feature processing method integrating multi⁃layer CNN components and attention mechanism was proposed,which parallelly encodes multiple features of input speech data such as Mel frequency cepstral coefficients,waveform features,speech spectrograms and analyzes the importance of these features through attention mechanism.The experimental results show that the performance of the proposed speech satisfaction recognition model is significantly better than traditional deep learning algorithms,with an accuracy of 98.31%,which is 5.58%higher than the best performing BiLSTM.
作者
钱旭盛
何玮
康雨萌
吴明
周玉
QIAN Xusheng;HE Wei;KANG Yumeng;WU Ming;ZHOU Yu(State Grid Jiangsu Marketing Service Center,Nanjing 210000,China)
出处
《电子设计工程》
2025年第24期178-182,共5页
Electronic Design Engineering
基金
国网江苏省电力有限公司科技项目(J2021151)。
关键词
电力智能客服系统
卷积神经网络
注意力机制
语音识别
power intelligent customer service system
Convolutional Neural Network
Attention Mecha⁃nism
speech recognition
