In order to improve the efficiency of emergency response and optimize the utilization of emergency resources, domain experts actively create role network of emergency response. Creating emergency role network manually is a time-consuming and labor-intensive task. An approach to extracting role networks from emergency plans is proposed, and then the extracted role relation network is made a quantitative analysis in this paper. First, role relation network is generated from emergency plan, which includes the emergency department and personnel are identified through a Bi-LSTM-CRF network, coreference resolution is implemented based on RoBERTa-E2E-Coref model, role relation is extracted based on the RoBRETa-CasRel model, and role relation network of emergency response is generated based on identified roles and their relationship. Second, role relation network from emergency plan with different levels were analyzed quantitatively by using the complex network analysis methods. Finally, experiment evaluation and a case study are given based on the real data sets, and the results show that the proposed approach can be used to assist emergency decision-makers to create role relation networks of emergency response. In addition, through the analysis of the extracted role relationship network, the quality of the emergency plan can be indirectly reflected.