基于NLP模型的智能IVR语音呼叫系统设计  

Design of Intelligent IVR Voice Call System Based on NLP Model

作　　者：韦国惠 王利超 钟世文 黄绪荣 李姗珊 Wei Guohui;Wang Lichao;Zhong Shiwen;Huang Xurong;Li Shanshan(Guangxi Power Grid Company Limited,Nanning 530023,China)

机构地区：[1]广西电网有限责任公司,南宁530023

出　　处：《单片机与嵌入式系统应用》2023年第10期69-73,共5页Microcontrollers & Embedded Systems

摘　　要：设计了一种基于NLP模型的智能IVR语音呼叫系统。在硬件方面,完成了无线呼叫器和显示终端设计。在软件方面,首先建立了多信道协议通信模式,运用优先级理念实现呼入语音接纳控制,然后搭建了IVR交互序列,将通信语音转入序列中,通过语音平台、解析器、网页服务器进行交互处理。最后以引入对抗训练算法的NLP模型为核心设计了一种可用于智能呼叫问答的预训练模型,在交互过程中不断学习,搜索出与用户呼叫语音相对应的处理方案,并通过语音播报、提示灯闪烁等方式提醒工作人员处理语音呼叫问题。系统测试结果表明,在不同的呼叫距离下,系统语音呼叫响应延时始终低于1 ms,满足呼叫响应的实时要求。In the paper,an intelligent IVR voice call system based on the NLP model is designed.In terms of hardware,the design of wireless pagers and display terminals are completed.In terms of software,a multi-channel protocol communication mode is established,use the priority concept to achieve incoming voice admission control,and then build an IVR interaction sequence to transfer the communication voice into the sequence for interactive processing through the voice platform,parser,and web server.Finally,with the introduction of adversarial training algorithms in the NLP model as the core,a pre-training model that can be used for intelligent call Q&A is designed.During the interaction process,continuous learning is carried out to search for processing solutions corresponding to user call voice,and staff are reminded to handle voice call problems through methods such as voice broadcasting and flashing warning lights.The system test results show that under different call distances,the system's voice call response delay is always less than 1 ms,meeting the real-time requirements of call response.

关 键 词：NLP模型 交互式语音响应 语音呼叫 

分 类 号：TN912.34[电子电信—通信与信息系统]