Remote supervision relation extraction method of power safety regulations knowledge graph based on ResPCNN-ATT

In the field of electric power, there is a scarcity of data related to the entity relationship of power safety regulations, and the types of entity relationships in the text data of power safety regulations are too complicated, and the task of extracting the relationships between power safety entities is relatively complex. In response to the above problems, this paper proposes a remote supervision relationship extraction method based on deep residual learning (Res) combined with multi-level attention mechanism. Firstly, according to the word vector and word position vector as input, the PCNN model is used to extract the semantic features of the text, and the deep residual is used to learn less the influence of noise data to better extract the deep semantic features of the power safety text sentence. Secondly, the multi-level attention mechanism is used to calculate the correlation between the corresponding entity and the context word, so as to assign different weights to different entity features and reduce the weight of noise data. Finally, through the attention mechanism of the relationship layer, it automatically learns the dependency and inclusion relationships between different relationships, and uses the softmax function to predict entity relationships. The analysis of simulation examples proves the effectiveness of the method in this paper.