In order to tackle the challenges of uneven entity distribution and fuzzy boundaries in fire accident information extraction, a fire information extraction model RA-BMAF (residual attention-based multi-head fusion) based on adversarial training and multi-head attention mechanism was proposed. A total of 264 fire accident reports were used as samples for entity labeling, and fire text test data were constructed. The RoBERTa-wwn-ext model was employed to generate dynamic word vectors, and adversarial samples were generated by perturbing these vectors through adversarial training. These samples were then co-embedded with the original inputs for joint training, thereby alleviating the problem of entity distribution imbalance in fire samples. Additionally, in the feature extraction layer, the multi-head attention mechanism was incorporated to further optimize the weight distribution of the sequence features extracted by the BiGRU (bidirectional gated recurrent unit) model and to enhance focus on the fuzzy boundaries of entities. In the decoding layer, conditional random fields were used to predict the sequence features. The results show that the model achieved precision, recall, and F1 scores of 83.10%, 91.13%, and 86.89%, respectively, outperforming the RoBERTa-wwn-ext-BiGRU-CRF (conditional random field) baseline model. It is concluded that the proposed approach significantly enhances the entity extraction process by addressing both entity imbalance and fuzzy boundaries in fire accident reports.
ZHAO Bao-xi, WANG Ming-da, LENG Gao-qiang.
Fire Accident Information Extraction Method Based on Adversarial Training and Multi-head Attention Mechanism[J].
Science Technology and Engineering, 2026, 26(3): 1307-1313 DOI:10.12404/j.issn.1671-1815.2501485
为了更好地表示获取到的中文文本的输入特征,使用RoBERTa-wwn-ext预训练模型火灾事故报告进行特征提取。相比于BERT,在预训练阶段使用了基于中文的全词掩码(whole word mask,WWM)技术,能够处理更复杂的上下文信息和语义关系。该模型由12层Transformer堆叠而成,每层包含768个隐层单元。
HanischD, FundelK, MevissenH T, et al. ProMiner: rule-based protein and gene entity recognition[J]. BMC Bioinformatics, 2005, 6: 1-9.
[2]
QuimbayaA P, MuneraA S, RiveraR A G, et al. Named entity recognition over electronic health records through a combined dictionary-based approach[J]. Procedia Computer Science, 2016, 100: 55-61.
[3]
SrivastavaS, PaulB, GuptaD. Study of word embeddings for enhanced cyber security named entity recognition[J]. Procedia Computer Science, 2023, 218: 449-460.
[4]
LiD Y, LiY, YangJ Z, et al. Dependency syntax guided BERT-BiLSTM-GAM-CRF for Chinese NER[J]. Expert Systems with Applications, 2022, 196: 116682.
[5]
HuangY, LiY, ZhangX. DTDM: dynamic temporal convolutional network and dynamic multihead attention for Chinese named entity recognition[J]. IEEE Access, 2023, 11: 128153-128161.
ZhangJiyuan, QianYurong, LengHongyong, et al. Survey of named entity recognition research based on deep learnin[J]. Modern Electronics Technique, 2024, 47(6): 32-42.
PuPan, ZhangYue, LiuYong, et al. Transformer optimization and application in named entity recognition of apple diseases and pests[J]. Transaction of the Chinese Society for Agricultural Machinery, 2023, 54(6): 264-271.
ChenJindong, HuChao, HaoLingxiao, et al. Named entity recognition for clothing quality sampling notice based on BERT-TENER[J]. Science Technology and Engineering, 2024, 24(34): 14754-14764.
ChengQuan, ZhangShuangbao. Research on feature-enhanced model for entity recognition of safety accident text based on deep leaning[J]. Journal of Safety Science and Technology, 2024, 20(6): 58-66.
WangMingda, ZhangBang, WuZhisheng, et al. A method for information extraction of urban gas accidents based on reinforcement learning[J]. Journal of Safety Science and Technology, 2023, 19(3): 39-45.
[18]
DevlinJ. Bert: pre-training of deep bidirectional transformers for language understanding[J]. arXiv Preprint, arXiv: 1810. 04805, 2018.
HaoKuangong, DongBing, WuYue, et al. Airport class based on BERT-BiLSTM-CRF model Chinese navigation notice element entity recognition[J]. Science Technology and Engineering, 2024, 24(10): 4182-4188.
XuYaxi, MengTianyu, WangXin, et al. Named entity recognition of aviation unsafe events embedded with fusion domain dictionary[J]. Science Technology and Engineering, 2024, 24(8) : 3284-3290.
XiaZhanjie, ZhangBeike, GaoDong. Named entity recognition of HSE inspection minutes based on data enhancement[J]. China Safety Science Journal, 2022, 32(12): 53-62.
GuanSiqi, DongTingting, WanZijing, et al. Fire accident case named entity recognition based on BERT-CRF model[J]. Fire Science and Technology, 2023, 42(11): 1529-1534.
[30]
TeamG L M. ChatGLM: a family of large language models from GLM-130B to GLM-4 all tools[J]. arXiv Preprint, arXiv: 2406. 12793, 2024.
[31]
LaffertyJ, McCallumA, PereiraF. Conditional random fields: probabilistic models for segmenting and labeling sequence data[C]// Proceedings of the Eighteenth International Conference on Machine Learning. San Francisco: Morgan Kaufmann Publishers Inc., 2001: 282-289.