HAZOP Text Named Entity Recognition using CNN-BilSTM-CRF Model

Hazard and Operability Study (HAZOP) is a Hazard analysis method based on system engineering, which analyzes the potential risks that may exist in the process of technological operation through a certain joint, and provides guidance for subsequent control and management. However, for a long time, the form of HAZOP analysis results is difficult to unify. In order to solving this problem, information needs to be extracted to establish a database, and Named Entity Recognition (NER) is a key step in this process. In this paper, a CNN-Bilstm-CRF neural network model is proposed for entity recognition of HAZOP text. A novel activation function is used in the convolutional neural network (CNN) for parameter optimization, and different weight initialization methods are selected to optimize the results. The corpus used in this paper is the HAZOP analysis report of the oil synthesis equipment of the self-marked coal seam indirect liquefaction project. Verified on the Tensorflow, the experimental results show that this method has a good recognition effect on HAZOP text entities, and the F1 value reaches 91.11%.