Accidents in chemical production usually result in fatal injuries, economic losses, and negative social impacts. To ensure personnel security in such cases, previous research has often used digital data, such as physical signals. However, the valuable textual information contained in chemical security texts, such as expert knowledge, has not yet been explored. Therefore, there is an increasing demand to mine useful information from these unstructured data. In this study, natural language processing (NLP) was applied to the hazard and operability (HAZOP) analysis reports. The classification model was trained to learn the classification of consequence severity levels in high-quality HAZOP analysis reports, which will not only ensure the consistency of the analysis results, but also help smaller chemical plants perform security analysis. In the classification model, we introduced Bidirectional Encoder Representation from Transformers (BERT), which, for word embedding, which is a powerful NLP pre-training model and significantly improved the effectiveness of the model. Through these application scenarios, the feasibility and possibility of applying NLP in chemical security text have been confirmed to a certain extent. In addition to digital data, future security managers will be able to monitor chemical production using natural language.

化学品生产中的事故通常会导致致命伤害、经济损失和负面社会影响。在这种情况下，为了确保人员安全，以前的研究经常使用数字数据，例如物理信号。然而，化学安全文本中包含的有价值的文本信息，如专家知识，尚未得到探索。因此，从这些非结构化数据中挖掘有用信息的需求越来越大。在这项研究中，自然语言处理（NLP）被应用于危害和可操作性（HAZOP）分析报告。训练分类模型学习高质量HAZOP分析报告中后果严重程度的分类，不仅可以保证分析结果的一致性，还可以帮助较小的化工厂进行安全分析。在分类模型中，我们引入了 Bidirectional Encoder Representation from Transformers (BERT)，对于词嵌入，这是一个强大的 NLP 预训练模型，显着提高了模型的有效性。通过这些应用场景，在一定程度上证实了NLP在化学安全文本中应用的可行性和可能性。除了数字数据，未来的安全管理人员将能够使用自然语言监控化学品生产。在化学安全文本中应用NLP的可行性和可能性已经得到一定程度的证实。除了数字数据，未来的安全管理人员将能够使用自然语言监控化学品生产。在化学安全文本中应用NLP的可行性和可能性已经得到一定程度的证实。除了数字数据，未来的安全管理人员将能够使用自然语言监控化学品生产。