An Entity Recognition Approach of GIS Fault Text

Entity recognition of GIS fault text contributes to the manual experience sharing and the execution efficiency of GIS operation and maintenance. However, traditional models have poor performance on such entity recognition tasks. In this paper, a novel entity recognition model is proposed. The model consists of several single-type entity recognizers, which increase the accuracy of the model. Bidirectional Encoder Representation from Transformers (BERT) is also applied to understand the characters in the text. Experiments show that the proposed model can achieve F1-score of 88.0%. Compared with other traditional entity recognition models, the performance of the proposed model is improved by about 20%.