Robot Intent Recognition Method Based on State Grid Business Office

Artificial intelligence is currently in an era of change, not only changing the artificial intelligence technology itself, but also changing human society. It has become more and more common to use artificial intelligence as the core human-computer interaction technology to replace manpower. Intention recognition is an important part of the human-machine dialogue system, and deep learning technology is gradually being applied to the task of intent recognition. However, intent recognition based on deep learning often has problems such as low recognition accuracy and slow recognition speed. In response to these problems, this paper designs a BERT fine-tuning to improve the network structure based on the pre-training model and proposes new continuous pre-training goals. To improve the accuracy of intent recognition, a method based on multi-teacher model compression is proposed to compress the pre-training model, which reduces the time consumption of model inference.