Parameter-Based Optimization Approach for Chinese Named Entity Recognition in Navy Equipment Unmanaged System

With continuous updates and replacements of naval equipment in various countries around the world, a large amount of text data containing essential entity information has been accumulated, possessing significant research and application value. However, due to the high confidentiality of naval equipment information, publicly available data, and mature knowledge graph construction methods specific to this vertical domain are scarce. Furthermore, the parameter settings for Chinese Named Entity Recognition (NER) techniques mainly rely on experience and subjective methods currently. To solve these issues, this paper collects publicly available information to build a dataset for the naval equipment domain and proposes a Dynamic Particle Swarm Optimization algorithm (D-PSO) for optimizing parameters in the BERT-BiLSTM-CRF naval equipment NER method. The research results demonstrate that the dynamic learning factors and dynamic weighting strategies in the proposed D-PSO algorithm significantly enhance the training efficiency of the DPSO+BERT-BiLSTM-CRF model, resulting in precision, recall, and F1 score reaching 90.66%, 92.08%, and 91.36%, respectively. This fully validates the advantages of the proposed improved model in terms of accuracy, effectiveness, and feasibility.