Named Entity Recognition of Power Substation Knowledge Based on Transformer-BiLSTM-CRF Network
Qing Yang, Jing Jiang, Xinze Feng, Jun He, Bolin Chen, Zepeng Zhang
Abstract
Power substations have accumulated a large number of knowledge texts in various forms. Named Entity Recognition (NER) of power substation knowledge can identify entities from these texts and lays the foundation for the subsequent knowledge management. To realize the entity recognition of power substation knowledge more efficiently, an improved Transformer- BiLSTM-CRF model is proposed. The model consists of embedding layers, improved Transformer module and BiLSTM-CRF module. The proposed model is compared with LSTM, BiLSTM, CRF, Transformer and their combined models in the dataset provided by power enterprises. The result on the test dataset shows that the precision is 84.8%, the recall is 85.8%, and the F1-score is 0.853, which are all better than the comnarative models.