Research on human-machine interaction method of smart grid based on meta-learning

Abstract

As the size of the power grid continues to grow, the workload of dispatchers is becoming more and more powerful, and dispatchers alone cannot meet the daily dispatching communication needs. Voice human-machine interaction is a key technology in the field of artificial intelligence. This article proposes a smart grid human-machine interaction method based on meta-learning, this article proposes to use meta-learning to train language models and first-order MAML to optimize the initial model parameters. Experimental results show that after 20 epochs of training using less data, the speech recognition WER is reduced to 1.64% and the SER is reduced to 9.78%, which is better than the traditional hidden Markov model on WER and SER, and with the increase of training times, the recognition performance is gradually improved, and there is no overfitting phenomenon.