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.