Power grid risk text segmentation method based on BERT neural network

Shu-ping Xu; Han-qiang Liu

doi:10.1117/12.3015387

6 February 2024 Power grid risk text segmentation method based on BERT neural network

Shu-ping Xu, Han-qiang Liu

Proceedings Volume 12979, Ninth International Conference on Energy Materials and Electrical Engineering (ICEMEE 2023); 129796M (2024) https://doi.org/10.1117/12.3015387
Event: 9th International Conference on Energy Materials and Electrical Engineering (ICEMEE 2023), 2023, Guilin, China

Abstract

In order to realize efficient automatic segmentation of power grid risk text information, an improved BERT (Bidirectional Encoder Representation from Transformers) model of power grid risk text segmentation model is proposed. The text data was pre-trained by BERT, BIO labeling was performed, and word vectors were obtained by encoding single Chinese characters according to double-byte encoding. Then BILSTM-CRF model was introduced to classify text labels, and the classic self-attention layer in BERT model was replaced by ProbSparse sparse self-attention layer. The performance of the model remains unchanged while the complexity of the model is reduced. The segmentation performance test is carried out on the transmission network risk text data set. The experimental results show that the segmentation accuracy, recall rate and F1 value of this model are all about 95.8%, which has a good segmentation accuracy, and the accuracy is three to five percentage points higher than that of BILSTM-CRF, CNN+LSTM, CNN+CRF and other comparison models, respectively, and effectively improves the segmentation accuracy of power grid risk texts.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Shu-ping Xu and Han-qiang Liu "Power grid risk text segmentation method based on BERT neural network", Proc. SPIE 12979, Ninth International Conference on Energy Materials and Electrical Engineering (ICEMEE 2023), 129796M (6 February 2024); https://doi.org/10.1117/12.3015387

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available