Abstract: To realize the functions of intelligent semantic search, intelligent question answering, and visual decision-making in the field of coal mine personnel violations, this paper first establishes a named entity recognition model and a sequence annotation model for coal mine personnel violations based on BERT-BiLSTM-CRF. Then, a corpus for named entity recognition of coal mine personnel violations is constructed, and the corresponding named entity annotations are obtained. Finally, four models, namely Word2vec-LSTM, BiLSTM-CRF, BERT-CRF, and BERT-BiLSTM-CRF, are employed to conduct named entity recognition experiments. The precision, recall, accuracy, and F₁-score of the four models are obtained and compared. A deep learning-based named entity recognition model （BERT-BiLSTM-CRF） for the coal mine violation domain is proposed and verified on the self-built coal mine violation corpus. The results demonstrate that, compared with the Word2vec-LSTM model, the BERT-BiLSTM-CRF model presents obvious advantages in terms of precision, accuracy, recall, and F₁-score. The combination of BERT and BiLSTM contributes to a better recognition performance on named entities related to coal mine personnel violations. This study can provide theoretical support for the automatic construction of a knowledge graph in the field of coal mine personnel violations.