HAZOP entity-relation ternary recognition model based on deep learning

Industries such as chemical and manufacturing often have significant safety risks associated with their production processes, and it is critical to conduct a HAZOP safety analysis of the entire process before an accident occurs. However, the large amount of expert knowledge contained in HAZOP reports is not fully utilized. To solve the current problems, this paper proposes a fused deep learning model to extract unstructured expert knowledge of HAZOP by entity recognition and relation extraction of HAZOP safety analysis text. Compared with ordinary text, HAZOP text has a denser professional vocabulary, which leads to great difficulty in entity recognition. Based on the language feature of the HAZOP report, this paper combines the ACmix model, which is good at processing local features in the visual field, with the SPN model to effectively improve the recognition ability of the HAZOP entity-relation triplet. By changing the size of the convolution kernel and reducing the dimension of the attention mechanism, the ACmix model is applied to the text domain for the first time. The research results in this paper can be used in the future for the automatic construction of knowledge graphs in the fields of chemical industry and manufacturing, which can help enterprises analyze safety accidents as well as optimize and improve equipment.