A Power Grid Start-Up Work Ticket Event Extraction Method Based on Domain Knowledge Graph Enhanced Model
Konferenz: EMIE 2022 - The 2nd International Conference on Electronic Materials and Information Engineering
15.04.2022 - 17.04.2022 in Hangzhou, China
Tagungsband: EMIE 2022
Seiten: 5Sprache: EnglischTyp: PDF
Autoren:
Qiu, Longgang; Ping, Yuan; Jiang, Guangji; Huang, Shuai; Zhang, Yong; Deng, Min; Chen, Congtan (State Grid Huainan Electric Power Supply Company, Huainan, Anhui Province, China)
Zhang, Yue (NARI Group Corporation Co., Ltd., (State Grid Electric Power Research Institute Co., Ltd.,), Nanjing, Jiangsu Province, China & Beijing KeDong Electric Power Control System Co., Ltd., Haidian District, Beijing, China)
Zhang, Pei (School of Electrical Engineering, Beijing Jiaotong University, Beijing, China)
Rao, Guozheng (College of Intelligence and Computing, Tianjin University, Tianjin, China)
Inhalt:
The power grid start-up work ticket event extraction has excellent challenges in its text mining. However, exiting extraction models usually rely on hand-crafted features and ignore the domain knowledge. To better understand the context knowledge of the power grid start-up work ticket, we propose a power grid start-up work ticket event extraction method based on domain knowledge graph enhanced model. Our model has several features. First, our model is a joint end-toend model. It helps to jointly perform both trigger detection and argument detection with shared parameters. Second, we integrate CNN and Bi-LSTM-CRF to detect event triggers. Third, we propose Conceptual Context Tree-LSTM (CC Tree-LSTM) with the shortest path to detect event arguments. Multiple experiments have proved that the framework of this paper is superior to other comparison methods in the Precision, Recall, and F1 score of event extraction tasks. The Precision, Recall, and F1 score of our model reach 0.8032, 0.8127, and 0.8079. It verifies the feasibility and effectiveness of the method.