Key points are not available for this paper at this time.
Emergency-related knowledge from urban rail transit (URT) safety event case base is crucial for realizing intelligent emergency response in URT. However, safety event case records are non-standard and unstructured, leading to challenges in data mining, information induction, and knowledge organization. To obtain emergency-related knowledge effectively from URT safety event cases, this paper proposes a knowledge acquisition model incorporating label semantics and provides a knowledge acquisition degree evaluation method. In this study, the knowledge acquisition task is formulated as a machine reading comprehension (MRC) task. This formulation can take full advantage of the rich semantic information of labels, which can compensate for the drawbacks of the traditional sequence labeling model. The experiments conducted on the URT safety event case base demonstrate the effectiveness of the proposed methods. The knowledge acquisition model performs well on different labels and achieves a performance boost over baseline models. Note to Practitioners —Urban rail transit (URT) safety events are caused by various factors, such as natural disasters like earthquakes and floods, equipment failures, and terrorist attacks. et al. These events may disrupt the normal operation of URT, pose a great risk to personal safety, and result in property losses. URT safety event cases encompass essential emergency-related experience and knowledge, which serve as valuable references for managing new emergencies and mitigating potential losses. However, the record format of URT safety event cases is non-standard, with the majority stored as fragmented textual data, posing challenges in effectively extracting and utilizing the valuable knowledge they contain. Considering the data characteristics of URT safety event cases, three different methods are developed for knowledge acquisition. Among these methods, a knowledge acquisition model is proposed to extract emergency-related knowledge from URT safety event cases efficiently and accurately.
Zhu et al. (Mon,) studied this question.